Who this is for: Pharma and biotech IP teams, R&D leads, portfolio managers, and institutional investors tracking the $654 billion personalized medicine market. This guide covers patent eligibility post-Mayo, companion diagnostic IP, CDx commercialization, CRISPR patent thickets, AI inventorship doctrine, global prosecution strategy, and life cycle management. It is structured as a working reference, not a survey.
Top-Level Key Takeaways
Before diving into the detail, here is what this guide concludes:
The U.S. Supreme Court’s 2012 Mayo v. Prometheus ruling remains the single most disruptive force in personalized medicine IP. It did not merely raise the bar for diagnostic method patents. It reframed the entire question of what a company is actually patenting. Many diagnostic claims filed before 2012 were, functionally, claims on a correlation between a measurable metabolite and a clinical outcome. Mayo told patent holders that a correlation is a law of nature, and laws of nature are not yours to own.
The response from IP teams has been uneven. Some pivoted well, learning to claim the technical method of detection rather than the clinical inference drawn from it. Others spent years in expensive post-grant review proceedings before reaching the same conclusion. The gap in strategic sophistication, between the companies that internalized Mayo’s logic early and those that did not, now maps closely onto the gap in portfolio quality across the industry.
The second force reshaping the landscape is the convergence of companion diagnostics (CDx) with regulatory necessity. The FDA’s de facto policy of co-approving targeted drugs with their companion diagnostics has made CDx IP a first-order commercial asset, not an afterthought. The company that controls the approved diagnostic for a given drug controls the patient selection gateway. That gateway is worth more than many drug patents, yet it still receives far less IP investment than the therapeutic side.
CRISPR gene editing and AI-assisted drug discovery are not future concerns. They are creating patent prosecution problems now. The Broad Institute vs. UC Berkeley dispute over CRISPR in eukaryotic cells generated a thicket of overlapping, overlapping foundational patents that forces every therapeutic developer into an expensive licensing negotiation before Phase I. AI is raising the obviousness bar in ways patent examiners are only beginning to apply consistently, and the lack of any jurisdiction willing to recognize an AI system as an inventor creates documentation requirements that most R&D organizations are not yet meeting.
This guide addresses all of it, with specific prosecution tactics, commercial strategy frameworks, and IP valuation perspectives for each major technology area.
Section 1: The Market That Justifies the Investment
H2: Size, Growth, and Why the Blockbuster Model Is Losing
The global personalized medicine market stood at an estimated USD 654.46 billion in 2025 and is projected to reach USD 1,315.43 billion by 2034 at a compound annual growth rate of 8.1%. The more narrowly defined precision medicine segment, which excludes personalized nutrition and wellness products, is growing faster: projections show a CAGR of 16.5%, reaching USD 470.53 billion by 2034. A third data source from ResearchAndMarkets puts the broader market at USD 869.9 billion by 2030 at a CAGR of 8.5%.
These figures are important not because they establish that personalized medicine is large (everyone knows it is) but because they define the prize that IP strategy is protecting. A company that achieves ten years of meaningful exclusivity on a personalized cancer therapy in a market growing at 16% annually is protecting a materially different asset than the blockbuster drug IP of twenty years ago. The math changes the required investment in IP.
The oncology segment accounts for over 40% of the total market and has been the domain where precision therapy moved fastest from laboratory concept to commercial product. North America holds approximately 45% of the global market share, with the U.S. market alone valued at nearly USD 180 billion in 2024 and projected to pass USD 400 billion by 2034. The Asia-Pacific region is the fastest-growing geography, driven by government-funded precision medicine initiatives in China and South Korea, an aging population base, and a rapidly professionalizing IP enforcement infrastructure.
H3: The ‘Niche-Buster’ Economic Model
The shift away from blockbusters is not simply a scientific development. It is a structural change in how pharmaceutical revenue is generated, and it has direct consequences for IP architecture.
A traditional blockbuster serves the largest possible patient population. Its commercial logic depends on volume. The patent protection model built around that logic, a composition-of-matter patent on the active ingredient plus a few formulation and method-of-treatment patents, worked well when annual patient volumes were in the hundreds of thousands or millions.
A niche-buster targets a biomarker-defined sub-population. Volume is intrinsically limited. The economic model depends on high per-patient pricing, strong clinical differentiation, and durable market exclusivity. Drugs like Herceptin (trastuzumab), Gleevec (imatinib), and Kalydeco (ivacaftor) each exceeded USD 1 billion in annual revenue while serving patient populations measured in thousands to tens of thousands, not millions. That is only possible with premium pricing, and premium pricing is only defensible with strong patent protection. The IP architecture for a niche-buster therefore needs to be deeper and more multi-layered than for a traditional drug, because the commercial consequence of a single patent being invalidated is proportionally much larger.
H3: Market Fragmentation as a Competitive Weapon
Personalized medicine fragments markets. When a drug company characterizes a previously homogeneous patient population using a biomarker test, it is doing something strategically interesting: it is converting a large, heterogeneous market into several smaller, more tractable ones. Each sub-population then becomes a defensible franchise. A competitor who wants to enter that space must run its own trial in the same sub-population, replicate the biomarker identification, and navigate whatever IP surrounds the diagnostic test.
This fragmentation is both a scientific improvement and a competitive moat. IP teams that understand this dynamic build patent portfolios around each sub-population definition, not just around the drug molecule. The claim set matters enormously: claims that recite the biomarker-positive sub-population with specificity, define the diagnostic cutoff, and claim the method of treatment together create a much harder target for a generic filer than a composition-of-matter claim alone.
Key Takeaways: Section 1
The personalized medicine market is a USD 654 billion opportunity growing at 8 to 16 percent annually depending on the segment. The niche-buster economic model makes durable patent protection more critical, not less, compared to blockbuster-era drugs. Market fragmentation through biomarker stratification is both a scientific tool and a competitive barrier. IP architecture must reflect the niche-buster logic: deeper, more multi-layered, and anchored to specific patient sub-population definitions rather than just the drug molecule.
Section 2: The Patent Eligibility Problem — Mayo, Myriad, and the Law of Nature Trap
H2: Why § 101 Is the Dominant Legal Risk in Personalized Medicine
Section 101 of the U.S. Patent Act states that patents are available for ‘any new and useful process, machine, manufacture, or composition of matter.’ That language sounds expansive. Courts have spent decades narrowing it. The three judicial exceptions — laws of nature, natural phenomena, and abstract ideas — are the operative constraints for nearly every personalized medicine patent.
The core tension is structural. Personalized medicine is, at its scientific foundation, the identification of correlations between individual biological characteristics and clinical outcomes. A genetic mutation correlates with drug response. A metabolite level correlates with toxicity. A protein expression level correlates with survival. These correlations are discovered, not invented. And under U.S. patent law as interpreted since 2012, discovered correlations are laws of nature.
This does not mean diagnostic and personalized medicine patents are impossible. It means the claims must be drafted around what the inventor actually built, not around what the inventor found. The distinction is technically demanding, legally consequential, and often the difference between a valid, enforceable patent and a piece of paper that will be invalidated under inter partes review before it is ever asserted.
H3: Mayo Collaborative Services v. Prometheus Laboratories (2012) — The Ruling in Full
Prometheus held patents on methods of optimizing thiopurine drug dosing for autoimmune diseases. The claims covered: (1) administering thiopurine to a patient, and (2) measuring metabolite levels in the patient’s blood, with the claim language indicating that levels above a certain threshold suggested decreasing the dose and levels below another threshold suggested increasing it.
The Supreme Court found these claims patent-ineligible unanimously. Justice Breyer’s opinion identified the correlation between metabolite concentration and drug efficacy or toxicity as a natural relationship — a law of nature. The question then became whether the claim added an ‘inventive concept’ sufficient to transform the natural law into a patentable application.
The Court said no. The ‘administering’ step was routine clinical practice. The ‘determining’ step used conventional techniques. The ‘wherein’ clauses that identified the relevant metabolite thresholds were simply instructions to recognize the natural correlation. Taken together, the claims added nothing beyond the natural law itself.
The Mayo two-step framework that emerged from this ruling has been applied to every subsequent diagnostic method challenge:
Step 1: Is the claim directed to a law of nature, natural phenomenon, or abstract idea? Step 2: If yes, does the claim add something more — an ‘inventive concept’ — that is not routine, conventional, or well-understood?
Failing Step 2 does not require that each individual claim element be conventional. The court looks at the claim elements ‘as an ordered combination’ to determine whether they add something inventive beyond the judicial exception itself. This holistic analysis has created enormous uncertainty, because what counts as ‘inventive’ versus ‘well-understood’ is not a bright-line test. It is a judgment call that has produced inconsistent results across district courts and at the Federal Circuit.
H3: IP Valuation Impact: What Mayo Did to Diagnostic Patent Portfolios
For diagnostic companies with large pre-2012 patent portfolios, Mayo was a valuation shock. Many of those patents were structured exactly as Prometheus’s were: find a biomarker, claim the step of measuring it, and claim the clinical inference drawn from that measurement. Under the Mayo framework, the measurement step is typically conventional, and the clinical inference step is a direction to apply a natural law. These claims are now high-risk assets.
Any M&A transaction, licensing negotiation, or equity raise that involves a diagnostic patent portfolio built before 2012 requires explicit Mayo analysis. Acquirers should apply a systematic haircut to pre-2012 diagnostic method claims absent evidence of successful post-grant review or a Federal Circuit affirmance. The claims that survive Mayo analysis are those that embed a specific technical method of detection that was non-conventional at the time of filing, or that claim a novel combination of biomarkers identified through a non-obvious analytical process.
The secondary market for these patents has contracted substantially. Trade secret protection for LDT-based diagnostics has increased proportionally, for reasons discussed in Section 4.
H3: Association for Molecular Pathology v. Myriad Genetics (2013) — Gene Patents After the Ruling
Myriad Genetics discovered the precise location and sequence of the BRCA1 and BRCA2 genes and filed patents giving it exclusive rights to isolate those genes and use them for diagnostic testing. For years it was the sole U.S. provider of BRCA testing at a price of several thousand dollars per test.
The Supreme Court held that naturally occurring isolated DNA segments are products of nature and not patent-eligible, regardless of the effort required to isolate them. The act of isolation does not change the informational content of the gene, which is what makes it scientifically and clinically useful. The Court reasoned that Myriad found the BRCA genes; it did not create them.
The Court drew a clean line on cDNA. Complementary DNA is synthesized in a laboratory from an mRNA template with introns removed. The resulting molecule does not exist in nature. It is man-made and qualifies as patentable subject matter under Section 101. This distinction preserved the patentability of the engineered genetic constructs at the heart of gene therapy, mRNA vaccines, and CRISPR delivery systems.
The practical consequences of Myriad were immediate. Within weeks of the ruling, multiple competing labs began offering BRCA testing, and prices dropped substantially. The ruling also freed academic researchers from the chilling effect of Myriad’s enforcement activity, which had restricted the development of improved testing methods.
For companies with natural gene sequence patents, Myriad eliminated a category of IP asset. For companies building on synthetic genetic constructs, it confirmed that their foundational IP remains protected. The ruling drew a clear line: if the molecule exists in nature, even in slightly modified or isolated form, it is not yours to patent.
H3: Ariosa Diagnostics v. Sequenom (Fed. Cir. 2015) — Mayo’s Most Damaging Application
The Sequenom case involved a genuinely transformative invention: a non-invasive prenatal test (NIPT) based on the discovery that cell-free fetal DNA (cffDNA) circulates in maternal plasma. This finding enabled prenatal screening for chromosomal abnormalities without the risk of amniocentesis. The clinical value was unambiguous.
The Federal Circuit nonetheless found the patent claims ineligible under Mayo. The existence of cffDNA in maternal plasma was a natural phenomenon. The steps of amplifying and detecting that DNA using PCR and other standard molecular biology techniques were conventional. The court concluded that the claims were directed to a natural phenomenon with only routine steps appended, and therefore failed Step 2 of the Mayo framework.
The decision drew a pointed concurrence from Judge Linn, who acknowledged he was bound by Mayo but found the result troubling. A genuinely novel discovery about a previously unknown biological phenomenon, coupled with a practical clinical method for exploiting it, was held ineligible for patent protection. The implication for the diagnostic industry was stark: scientific novelty and clinical utility are insufficient to establish patent eligibility if the technical steps of detection are conventional.
Sequenom shifted resource allocation within diagnostic R&D organizations. Investment in developing proprietary, non-conventional detection methods increased, because those methods represent the only reliable route to a patent-eligible claim post-Ariosa. Investment in pure biomarker discovery research, where the discovery would be disclosed in a patent application that then failed eligibility, decreased or moved toward trade secret protection.
H3: The Alice Corp. Spillover Into Diagnostic Software
The Supreme Court’s 2014 ruling in Alice Corp. v. CLS Bank International extended the Mayo framework to abstract ideas implemented on a computer. For personalized medicine companies, this matters because modern diagnostic platforms increasingly rely on proprietary algorithms to interpret complex genomic, proteomic, or imaging data. If that algorithm is patented as a software claim and the underlying data analysis is characterized as an abstract idea, the claim faces an Alice challenge.
The Vanda Pharmaceuticals v. West-Ward Pharmaceuticals Federal Circuit decision in 2018 offered some relief. The court distinguished between diagnostic method claims directed to a natural law (as in Mayo) and method-of-treatment claims that apply a genetic-based insight to a specific treatment regimen. The claim in Vanda specified a particular dosage of iloperidone based on CYP2D6 genotype, and the court found it patent-eligible because it was directed to treatment, not to the observation of a natural correlation per se. The distinction is thin and the subject of ongoing academic and practitioner debate, but Vanda gives treatment-oriented personalized medicine claims a defensible pathway.
Key Takeaways: Section 2
Mayo is not merely a legal precedent; it is a structural constraint on the IP value of diagnostic biomarker discoveries. Pre-2012 diagnostic method patents require explicit eligibility analysis before any transaction. cDNA and synthetic genetic constructs remain patentable. The Sequenom outcome means that a diagnostic innovation built solely on a novel biomarker discovery, with conventional detection steps, will not be patent-eligible in the U.S. Claims must embed a non-conventional technical element, or the protection strategy must shift to trade secrets or the composition of matter (the diagnostic kit itself). The Vanda ruling provides a pathway for method-of-treatment claims tied to genotype-guided dosing, but that pathway is not broad and requires careful claim construction.
Investment Strategy: Section 2
Portfolio managers evaluating diagnostic companies should score pre-2012 diagnostic method claims at a substantial discount to face value absent a favorable Federal Circuit or Supreme Court ruling on eligibility. Companies with patent estates built around CDx kit components, proprietary detection chemistries, or novel algorithm-plus-treatment claims have structurally stronger IP positions than those relying on biomarker-correlation method claims. M&A due diligence should include a systematic Mayo/Alice audit of the target’s diagnostic patent estate with independent outside counsel confirmation, not just management representation.
Section 3: Novelty and Non-Obviousness in the Biomarker Age
H2: The Rising Evidentiary Bar for Personalized Medicine Patents
Even when a personalized medicine claim clears the Section 101 eligibility hurdle, it must independently satisfy novelty (35 U.S.C. § 102) and non-obviousness (35 U.S.C. § 103). Both requirements present distinct challenges in a field defined by massive public genomic databases, powerful computational tools, and the rapid expansion of published biomarker research.
H3: Sub-Population Novelty — The EPO’s Approach vs. the U.S.
A recurring scenario in personalized medicine prosecution is the following: a drug has been approved for a broad indication. A company later discovers, through retrospective analysis of clinical trial data, that the drug is highly effective in a biomarker-defined sub-population that was present in the original study but not characterized. The company files a patent claiming the method of treating that sub-population.
The novelty problem is that the broad prior approval necessarily encompassed patients with and without the biomarker. Some portion of those patients, therefore, received the drug before the patent application was filed. Were they the same as the newly claimed sub-population? Technically, yes. The members of the sub-population were treated, even if no one knew it at the time.
The European Patent Office has developed a workable doctrine here. Under EPO case law, a claim to a new therapeutic use in a defined patient sub-population can be novel if the sub-population is identified by a specific, measurable physiological or pathological characteristic, not merely by a statement of therapeutic efficacy. The claim cannot simply say ‘patients who respond to the drug.’ It must identify the biomarker, its detection method, and the clinical threshold that defines the group. This ‘clinical situation novelty’ doctrine requires precise technical characterization and rewards companies that invest in rigorous biomarker validation before filing.
U.S. practice does not have an equivalent formal doctrine, but similar logic applies through the enablement and written description requirements. A claim that precisely characterizes a biomarker-defined sub-population with measurable diagnostic criteria is both more defensible on novelty grounds and more likely to satisfy Section 112 than one relying on vague patient response language.
H3: Non-Obviousness and the Computational Skill Problem
The non-obviousness inquiry asks whether the differences between the claimed invention and the prior art would have been obvious to a person having ordinary skill in the art (PHOSITA) at the time of filing. As machine learning tools have become standard in genomics and bioinformatics labs, the capabilities of the PHOSITA have increased substantially. This is creating a rising bar for non-obviousness that is only beginning to be reflected in patent examination practice.
Consider a biomarker discovery made in 2024 by training a gradient boosting classifier on a publicly available TCGA cancer genomics dataset. If the biological marker identified was present in the public data, and the analytical method used was a published technique, a well-motivated examiner could argue that the discovery was the predictable result of applying known methods to publicly available data. That argument is not the same as saying the discovery was obvious before it was made, but it captures something real about the state of the art in bioinformatics. The combination of accessible data and accessible computational tools is compressing the window in which biomarker discoveries are genuinely non-obvious.
Companies filing biomarker patents in 2025 should anticipate this challenge and build a record of unexpected results. Unexpected results remain one of the strongest arguments for non-obviousness under U.S. case law. If the biomarker performs better than a skilled researcher would have predicted from the prior art, document that gap quantitatively, with comparisons to prior art performance benchmarks, not qualitative assertions of surprise. PTAB has shown consistent skepticism of non-obviousness arguments based on unexpected results when the results are not compared to the closest prior art with specific numerical data.
H3: Written Description, Enablement, and the Breadth Problem
Patent claims in personalized medicine often attempt to cover broad classes of biomarkers, antibodies, or genetic targets. The Supreme Court’s 2023 Amgen v. Sanofi ruling significantly tightened the enablement requirement for functional antibody claims. The Court held that a claim covering all antibodies that bind a given target and achieve a specific functional result must enable the full scope of what is claimed. A handful of examples, even several dozen, may be insufficient to enable a claim that functionally covers potentially millions of antibody variants.
The practical consequence for personalized medicine patent drafters is that broad functional claims on antibody-based diagnostics or therapeutics now carry higher prosecution and litigation risk. The application must disclose enough structural and functional data to allow a PHOSITA to make and use the full claimed scope without undue experimentation. Applicants should either narrow claims to reflect the specific examples disclosed with structural particularity, or invest in generating a larger and more structurally diverse set of working examples before filing. Filing broadly and hoping to narrow later, a common historical practice, now carries more risk given post-Amgen examination and PTAB practice.
Key Takeaways: Section 3
Sub-population claims can be prosecuted to novelty in the U.S. and Europe if the patient sub-population is defined by a specific, measurable biomarker with a disclosed diagnostic threshold. Non-obviousness is under increasing pressure from the expanding computational capabilities attributed to the PHOSITA; unexpected results arguments must be supported by quantitative comparisons to closest prior art. Post-Amgen, broad functional antibody claims on diagnostic or therapeutic biologics carry meaningful enablement risk and should be supported by structurally diverse working examples at filing, not just conceptual breadth.
Section 4: IP Architecture — Building a Portfolio That Holds
H2: The ‘Web’ Model vs. the Picket Fence
The traditional pharmaceutical IP strategy, a composition-of-matter patent on the active molecule plus a cluster of formulation and dosing method patents, was designed for small molecule drugs with a single active ingredient and a broad indication. It provided adequate protection when a single strong composition-of-matter patent could block all generic entry until expiration.
Personalized medicine requires a different architecture. The ‘web’ model acknowledges that no single patent will protect a personalized therapy franchise. Instead, the goal is a dense network of claims, each protecting a different layer of the product, such that a competitor who invalidates or designs around one layer is still blocked by several others. The layers of a well-constructed web include the drug molecule itself, the companion diagnostic kit and its key components, the method of treatment in a biomarker-defined population, the specific dosing regimen in that population, the manufacturing process for any biologic component, the software algorithm used to interpret diagnostic results, and any de novo patient selection criteria identified post-approval.
This is not just a litigation strategy. It reflects the commercial reality that a personalized medicine franchise has multiple distinct value-generating components, each of which can be independently attacked, licensed, or acquired. Managing them as separate assets within a unified strategic framework requires deliberate portfolio architecture, not opportunistic filing around whatever the chemistry team produces.
H3: Composition of Matter vs. Method Claims — A Practical Decision Framework
For personalized medicine IP teams, the choice between composition-of-matter and method claims is not binary; it is sequential and strategic.
Composition-of-matter claims on a novel synthetic molecule, a unique formulation, an engineered antibody with a defined structure, or a synthetic cDNA construct remain the most robust form of pharmaceutical patent protection available. They cover all uses of the claimed composition and are not subject to the Mayo law-of-nature problem in the same way method claims are. When the innovation includes a novel molecule or synthetic construct, composition claims should be the primary protection strategy.
Method-of-treatment claims in a biomarker-defined population add a second layer. They are potentially subject to Mayo challenges, but Vanda and subsequent Federal Circuit precedent provide a pathway when the claim is directed to treating a patient rather than to the natural correlation itself. These claims should specify the diagnostic step, the biomarker threshold, and the treatment protocol together in the claim language, rather than separating them into independent claims that can more easily be characterized as directed to a natural law.
Method-of-manufacture claims protect the manufacturing process, particularly for biologics where the process defines the product. These claims are generally not subject to Mayo challenges and become increasingly important as biologic manufacturing know-how becomes a key source of competitive differentiation.
Diagnostic kit claims, covering the physical components of the companion diagnostic test, its specific reagents, and its physical architecture, provide protection that is not subject to the same eligibility challenges as the method of diagnosis. The kit is a composition of matter. Claiming the specific combination of reagents, calibrators, and detection components used in the CDx creates a barrier that is independent of the method-of-diagnosis problem.
H3: Trade Secrets as a Complement, Not a Substitute
Post-Mayo, trade secret protection for diagnostic methods has increased substantially, particularly for laboratory-developed tests (LDTs) performed in-house. An LDT is a diagnostic test designed, developed, and performed within a single laboratory. The test methodology, including any proprietary algorithm used to score the result, can be maintained as a trade secret indefinitely, because the test is never disclosed to the public in the course of its use. The patient receives a result; the method by which that result was derived remains proprietary.
This is genuinely attractive under the post-Mayo framework. A diagnostic method that would be unpatentable, because it applies a natural correlation using conventional laboratory steps, may be highly valuable as a trade secret if the laboratory execution is sufficiently complex that competitors cannot easily replicate it. The risk is the classic trade secret vulnerability: reverse engineering, employee departure, and the possibility that a competitor independently develops the same method through their own research.
For high-throughput genomic diagnostics with complex multi-biomarker scoring algorithms, the trade secret model is viable and increasingly common. Companies like Foundation Medicine and Guardant Health have built significant competitive positions on diagnostics whose algorithmic cores are trade secrets rather than patents, combined with patent protection on the physical kit components and specific reagents. The hybrid model, patents on tangible components plus trade secrets on the algorithmic method, is the current state of the art for sophisticated diagnostic IP strategy.
H3: Evergreening in Personalized Medicine — The Technology Roadmap
Evergreening refers to the practice of filing follow-on patents designed to extend the effective exclusivity period of a drug franchise beyond the expiration of its original composition-of-matter patent. In traditional small molecule pharma, common evergreening tactics include new formulation patents, new dosing method patents, new indication patents, and new polymorph or salt form patents. Each extends exclusivity by some period beyond the original expiration.
In personalized medicine, evergreening has a richer toolkit because the product ecosystem is more complex.
The companion diagnostic itself generates a separate evergreening opportunity. The first approved CDx for a given drug creates a regulatory lock-in: the drug’s labeling specifies the diagnostic, and prescribers follow the labeled indication. A company that develops an improved diagnostic, one with better sensitivity, lower cost, or faster turnaround, and patents it, can potentially negotiate with the FDA for a label update that names the new test. If they own both the new diagnostic and the original drug, this generates a new exclusivity period on the diagnostic while reinforcing the drug’s franchise. If a competitor has developed the improved diagnostic, the original drug company faces the strategic decision of whether to partner or litigate.
Combination therapy patents represent a second evergreening vector specific to oncology. As treatment protocols evolve to combine targeted therapies with immunotherapies, a company that patents specific combination regimens in biomarker-defined patient populations can extend exclusivity beyond the expiration of either drug’s individual patents. The combination claim is patentable when it produces unexpected synergistic results that are documented with clinical data. The Keytruda combinations franchise at Merck illustrates this: by continuously generating clinical data for pembrolizumab in new combinations and new patient-defined sub-populations, Merck has built a combination patent estate that will support market exclusivity well beyond the expiration of the original PD-1 antibody patents.
New formulation patents for personalized biologics are a third vector. Subcutaneous formulations of drugs originally approved intravenously, fixed-dose combination products, and co-formulations with improved stability all generate new composition-of-matter patents. Roche’s subcutaneous trastuzumab (Herceptin SC) extended the HER2 franchise by years and captured premium reimbursement for the improved patient convenience.
Post-approval biomarker expansion patents represent a fourth and underutilized vector. When a drug is initially approved for a specific biomarker-positive population, subsequent clinical data often identifies additional biomarker sub-populations that also benefit. Filing method-of-treatment patents that cover these newly identified sub-populations, before the clinical data is published, creates additional exclusivity layers. These claims face novelty challenges if the drug was already being used in the broader population, so the EPO clinical-situation novelty doctrine and careful U.S. claim drafting around the specific biomarker combination are essential.
The complete evergreening technology roadmap for a personalized medicine product therefore includes: original composition claims (Year 0), biomarker-stratified method-of-treatment claims (Year 0 to 2), CDx kit claims (Year 1 to 3), formulation variant claims (Year 3 to 7), combination therapy claims (Year 3 to 10), follow-on indication claims in additional biomarker-defined populations (Year 5 to 12), and next-generation diagnostic claims (Year 7 to 15). Executed with discipline, this roadmap extends the period of meaningful commercial protection substantially beyond the 20-year life of the original composition patent.
Key Takeaways: Section 4
The web model of IP protection, layering composition, method, manufacturing, diagnostic, and algorithm claims, provides more durable exclusivity than any single patent. Composition-of-matter claims on synthetic constructs remain the most robust anchor. Trade secret protection is a viable complement for LDT-based diagnostic algorithms that would fail Mayo. Evergreening in personalized medicine has more tactical options than in small molecule pharma: CDx innovation, combination therapy patents, biomarker expansion, and formulation variants each generate new exclusivity layers when properly timed and executed.
Investment Strategy: Section 4
Assess the depth of a company’s IP web before valuing its exclusivity runway. A drug with a single composition patent expiring in three years and no companion diagnostic, combination, or formulation patents is a materially weaker asset than a drug whose underlying composition is off-patent but whose CDx, combination regimens, and next-generation formulation are all protected. The Herceptin franchise at Roche illustrates the valuation difference: trastuzumab generated over $20 billion in cumulative revenue largely through a combination of successive CDx, combination, and formulation patents rather than reliance on the original antibody composition claim alone.
Section 5: Companion Diagnostics — IP Strategy and Commercial Architecture
H2: Why the Diagnostic Is Worth More Than Its Revenue
The global companion diagnostics market was valued at approximately USD 7.6 billion in 2023 and is projected to reach USD 15.4 billion by 2028, a CAGR of 15.2%. These figures measure direct CDx revenue: kit sales, reagent contracts, and instrument placements. They substantially understate the strategic value of CDx IP, because the most important commercial function of a companion diagnostic is not what it earns directly but what it enables and protects.
A CDx patent portfolio performs four distinct functions in a personalized medicine franchise:
It controls the patient selection gateway. The approved CDx determines who receives the drug. This is not just a clinical function; it is a commercial one. Control of the patient selection process gives the CDx holder significant leverage over market access, formulary positioning, and competitive entry.
It creates regulatory lock-in. Once a drug is approved with a specific labeled companion diagnostic, that test becomes the standard of care. A competitor who wants to introduce a better diagnostic for the same drug must generate clinical evidence of equivalence or superiority and negotiate a label change, a process that typically takes three to five years and tens of millions of dollars. The first CDx is a structural barrier to entry.
It raises the NPV of the therapeutic program. Clinical models show that a CDx-enriched trial, by selecting likely responders, can increase the probability of trial success from around 20% to over 60% in oncology indications where a predictive biomarker exists. That improvement in probability of success, when discounted into a Net Present Value calculation, is worth hundreds of millions to billions of dollars. Published analyses have estimated NPV uplift of USD 1.8 billion from a successful CDx strategy in oncology, moving program NPV from USD 900 million to USD 2.7 billion.
It extends product lifecycle. A CDx that identifies a new responding sub-population post-approval can support a supplemental New Drug Application (sNDA) or supplemental Biologics License Application (sBLA) for a new indication in that sub-population, generating a new period of market exclusivity when combined with method-of-treatment patents on the new indication.
H3: FDA Regulatory Framework for Companion Diagnostics
The FDA formally defined the companion diagnostic in 2014. The agency’s guidance states that a CDx is an in vitro diagnostic device that provides information essential for the safe and effective use of a corresponding therapeutic product. The key operative word is ‘essential.’ When the FDA determines that a diagnostic is essential for safe and effective use, it will not approve the drug without contemporaneous approval or clearance of the diagnostic.
The FDA processes CDx approval through two pathways. The Premarket Approval (PMA) pathway applies to high-risk devices, which includes most CDx tests because of the clinical consequences of false positives or false negatives. The De Novo pathway applies to novel devices with moderate risk and no existing predicate. The 510(k) pathway, the standard route for lower-risk devices, is rarely used for CDx because of the high-risk classification of most companion diagnostics.
The co-development timeline has significant IP implications. A company that begins CDx development at Phase II, when biomarker hypotheses are first being tested, will file CDx patent applications significantly earlier than one that waits until Phase III data confirms the biomarker. Earlier filing means an earlier priority date, which means broader and more defensible claims in a crowded space. CDx IP strategy and CDx development timing are not separable decisions.
H3: CDx Business Models and Their IP Implications
The traditional CDx business model is the kit sale: the diagnostic company sells a reagent kit or a complete testing system to a hospital pathology department or reference laboratory, which performs the test and bills the payor. Revenue is transactional, driven by test volume.
More sophisticated models are emerging. The pay-per-reportable-result model charges the laboratory per valid clinical result rather than per kit consumed. This aligns revenue with clinical utility rather than with consumable volume and creates recurring, predictable revenue streams. Patenting the specific result-reporting methodology and the software interface through which results are delivered adds IP protection to this model.
The service-based or ‘closed system’ model involves the diagnostic company performing the test itself, either in a central laboratory or through a network of affiliated labs, rather than licensing the test methodology to third parties. This retains trade secret protection for the algorithmic core of complex multi-analyte tests, generates higher per-test margins, and enables continuous quality control. The trade-off is geographic coverage and the capital cost of maintaining testing capacity. Companies like Genomic Health (now part of Exact Sciences) built significant value on this model with Oncotype DX, a genomic test for breast cancer recurrence risk that is performed exclusively in Exact Sciences’ laboratory and whose algorithmic scoring system has never been disclosed in a patent application.
The partnership model, where the drug company and the diagnostic company co-develop the CDx under a formal agreement, requires explicit contractual treatment of IP ownership, royalty arrangements, and exclusivity. Contracts should specify who owns improvements to the CDx made post-approval, what happens to the diagnostic IP if the drug is acquired, and what territorial rights each party holds. Ambiguity in these agreements has generated significant commercial disputes. The Roche/Dako relationship for HercepTest, and the Merck/Dako relationship for the PD-L1 22C3 pharmDx assay, both required careful contractual architecture to allocate IP rights across multiple jurisdictions.
H3: The Biomarker Specificity Problem — When the CDx Is Wrong
A structural vulnerability in every personalized medicine franchise is the diagnostic itself. Biomarkers are rarely perfect. PD-L1 expression, the companion diagnostic biomarker for pembrolizumab in lung cancer, is measured using immunohistochemistry with a tumor proportion score cutoff. But PD-L1 expression is heterogeneous within tumors, can change over time, and does not perfectly predict which patients respond to checkpoint inhibitors. Some PD-L1-low patients respond; some PD-L1-high patients do not.
This imperfection is both a scientific problem and a competitive opportunity. A company that develops a more precise companion diagnostic, one that better discriminates responders from non-responders in a given indication, does not need to develop a new drug. It can seek a label expansion for an existing drug using the superior diagnostic. If that company holds patents on its improved CDx, it is building commercial value directly from the diagnostic, not from the therapeutic.
IP teams at personalized medicine companies should conduct systematic CDx performance reviews at regular intervals post-approval, comparing their approved CDx to emerging diagnostic technologies in the same biomarker space. Where competitive CDx innovation is advancing, the company faces a choice: acquire or license the improved diagnostic and integrate it into the franchise, or compete against it. Waiting is not a neutral option. A competitor with a superior CDx can gain prescribing share by demonstrating that more of its patients are true responders, even using the same drug.
Key Takeaways: Section 5
The companion diagnostic generates commercial value far exceeding its direct revenue. CDx IP controls patient selection, creates regulatory lock-in, increases therapeutic program NPV, and supports life cycle extension. FDA requires contemporaneous CDx approval for drugs that rely on the diagnostic for safe and effective use, making CDx IP on the critical path of the therapeutic program. The service-based CDx model protects algorithmic trade secrets while generating high per-test margins. Biomarker imperfection is a competitive vulnerability; IP teams should monitor CDx performance against emerging alternatives and act before competitors do.
Section 6: Case Studies in Precision IP — Herceptin, Keytruda, and Kalydeco
H2: What the Three Most Important Personalized Medicine Franchises Actually Did
These case studies are not historical anecdotes. They are working templates for IP and commercial strategy in the personalized medicine era. The decisions made by Genentech, Merck, and Vertex during the development of their flagship personalized medicines define the state of the art for the field.
H3: Herceptin (Trastuzumab) and HercepTest — The Pioneer Template
Genentech’s development of trastuzumab established every major structural element of the personalized medicine IP playbook two decades before the FDA formalized its CDx approval policy.
The scientific background: roughly 15 to 25 percent of breast cancers, specifically the range varies by staging and patient population, show amplification of the HER2 gene with consequent overexpression of the HER2 protein on the tumor cell surface. HER2-positive tumors are biologically more aggressive and associated with worse prognosis at diagnosis. They are also, critically, dependent on HER2 signaling for proliferation, making the protein an actionable therapeutic target.
Genentech generated trastuzumab, a monoclonal antibody that binds the extracellular domain of the HER2 protein and inhibits downstream signaling while also directing antibody-dependent cellular cytotoxicity against HER2-positive cells. The molecule is patentable as a synthetic antibody composition of matter, a synthetic biological construct not found in nature, and the composition patents were the anchor of the IP estate.
The strategic decision that defined the franchise was the early partnership with Dako (now Agilent) to co-develop the HercepTest immunohistochemistry assay. Genentech recognized that without a standardized diagnostic, HER2-positive patients would be identified inconsistently across institutions, trial results would be diluted by HER2-negative patients enrolled in error, and the drug’s efficacy signal would be obscured. The decision to invest in CDx development parallel to the clinical program, not after it, was the organizing principle of the entire commercial strategy.
The FDA approved trastuzumab and HercepTest on the same day in September 1998. This simultaneous approval was not planned by the FDA as a policy statement; it was the logical consequence of Genentech’s co-development timeline. But it established a precedent that reshaped how the agency thinks about targeted therapy approvals for the next 25 years.
IP valuation of the Herceptin franchise: Herceptin generated cumulative global revenue exceeding $60 billion from approval through patent expiration. The original antibody composition patents were the primary protection through the mid-2010s. Roche, which acquired Genentech, then executed a comprehensive evergreening program. Subcutaneous trastuzumab (Herceptin SC) received approval in Europe in 2013, generating new formulation patents. Trastuzumab emtansine (T-DM1, Kadcyla), an antibody-drug conjugate combining trastuzumab with a cytotoxic payload, generated entirely new composition patents with expiries extending well into the 2030s. Trastuzumab deruxtecan (T-DXd, Enhertu), developed in partnership with Daiichi Sankyo, introduced a third generation of HER2-targeted patent-protected assets. The franchise that began with a single antibody composition patent in 1998 has been successfully extended through CDx innovation, formulation modification, antibody-drug conjugate development, and partner collaboration for over 25 years, and the exclusivity runway continues.
H3: Keytruda (Pembrolizumab) and PD-L1 22C3 pharmDx — The Immuno-Oncology Template
Merck’s development of pembrolizumab, the world’s highest-grossing drug as of 2023 with revenues exceeding USD 25 billion annually, exemplifies the modern iterative CDx expansion strategy. Unlike Herceptin, where a single biomarker (HER2) defined a relatively stable patient population, Keytruda’s commercial strategy has involved continuous expansion across new biomarker-defined populations in dozens of tumor types.
The scientific mechanism: pembrolizumab is a humanized IgG4 monoclonal antibody that binds to the PD-1 receptor on T cells and blocks its interaction with PD-L1 and PD-L2 on tumor cells and antigen-presenting cells. This blockade prevents the immunosuppressive ‘off signal’ that tumor cells use to evade T cell recognition. The anti-tumor activity is not tumor-type-specific but depends on whether the patient’s immune system has been primed and whether the tumor expresses sufficient PD-L1 (or has high tumor mutational burden) to generate a response.
The 2015 accelerated approval in metastatic NSCLC with PD-L1 expression determined by the Dako PD-L1 IHC 22C3 pharmDx assay established the template: drug plus CDx plus specific biomarker threshold (TPS of 1% or greater for second-line; TPS of 50% or greater for first-line monotherapy approval in 2016). Each new approval has added a new CDx or CDx threshold that is patent-protected and labeled on the drug.
The IP architecture of the Keytruda franchise is deliberately multi-layered. The core PD-1 antibody composition patents are the anchor. A second ring of patents covers specific methods of treatment in PD-L1-defined sub-populations across different tumor types. A third ring covers combination therapy methods with specific agents (e.g., pembrolizumab plus chemotherapy, pembrolizumab plus VEGF inhibitors, pembrolizumab plus CTLA-4 inhibitors) in tumor-type and biomarker-defined populations. A fourth ring covers the CDx assay components, detection methods, and scoring algorithms. The result is that even as the core antibody patents expire, the combination therapy patents, indication-specific method patents, and CDx claims extend commercially meaningful exclusivity substantially beyond the composition patent expiry.
A critical strategic lesson from Keytruda’s 2028 PD-1 antibody composition patent expiry situation: Merck has been running and publishing data from hundreds of KEYNOTE trials for over a decade. These trials generate clinical data that is incorporated into new method-of-treatment patent claims filed annually. By the time the composition patent expires, there will be dozens of indication-specific and combination-specific method patents expiring at various dates through the mid-2030s. Biosimilar pembrolizumab manufacturers will be able to make the molecule but will face complex freedom-to-operate questions around every specific labeled indication. That is a planned outcome, not an accident.
H3: Kalydeco (Ivacaftor) — Genetics-First Drug Development and the CFTR IP Model
Vertex Pharmaceuticals’ development of ivacaftor for cystic fibrosis patients with specific CFTR gating mutations represents a different structural model: starting from a genetic target definition rather than a clinical disease category. This approach generates a naturally strong IP position because the target definition itself, the specific gating mutation class, is encoded in the patent claims from the outset.
Cystic fibrosis is caused by over 2,000 known mutations in the CFTR gene. Most mutations in the most common class (F508del, the most prevalent CFTR mutation) cause CFTR protein to misfold and fail to reach the cell surface. A different class of mutations, called gating mutations (including G551D, the most common), allows the protein to reach the surface but produces a channel that opens insufficiently. Ivacaftor is a CFTR potentiator: it binds to the channel and increases its open probability, restoring partial CFTR function.
Ivacaftor is only effective in patients with specific gating mutations, which represent approximately 4 to 5 percent of the cystic fibrosis population in the United States. A genetic test for CFTR mutation status is required before prescribing, and Vertex’s labeling specifies the mutation classes that qualify. The IP estate therefore includes: composition-of-matter patents on the ivacaftor molecule, method-of-treatment patents in patients with specific CFTR mutations, CDx-related claims on the genetic testing methodology, and combination patents covering ivacaftor in combination with lumacaftor (Orkambi) and tezacaftor (Symdeko) for other CFTR mutation classes.
The genetic-first development model has a specific IP advantage: the biomarker is precisely defined at the genetic sequence level, which creates crisp patent claim boundaries. There is no ambiguity about who is in the patient population. This eliminates the diagnostic threshold imprecision problems that affect PD-L1-based CDx and reduces the surface area for competitor CDx challenges. The mutation either is or is not present. That binary definition provides cleaner IP and cleaner commercial positioning.
IP valuation note: Vertex’s CFTR franchise generated USD 9.86 billion in revenue in 2023 from a patient population estimated at around 30,000 patients in the U.S. That is approximately USD 328,000 per patient per year. The economic logic depends entirely on the patent protection that supports this pricing, because no single-payor would sustain that cost for a generic therapy. The CFTR franchise illustrates the direct line between IP strength and the financial sustainability of ultra-rare-disease drug development.
Key Takeaways: Section 6
Herceptin established the co-development template and the simultaneous drug-CDx approval model. Its 25-year franchise extension through ADC development, formulation modification, and CDx innovation is the most complete example of personalized medicine evergreening in the industry. Keytruda’s strategy of continuous biomarker expansion, combination trial generation, and indication-specific method patent filing is creating an exclusivity structure that extends well beyond the antibody composition patent expiry through a deliberately accumulated portfolio of method and combination claims. Kalydeco demonstrates that genetics-first drug development, targeting a specific mutation class rather than a disease category, generates cleaner IP boundaries and supports extreme per-patient pricing when the patient population is small and the clinical benefit is large.
Investment Strategy: Section 6
For portfolio managers, the Herceptin and Keytruda models suggest that the long-term revenue sustainability of a personalized medicine franchise is more accurately measured by the depth of the IP web than by the remaining life of the primary composition patent. Analysts who forecast revenue cliffs based solely on primary patent expiry will systematically undervalue franchises with active CDx innovation programs, combination trial pipelines, and continuously filed method-of-treatment patent applications. Build a Keytruda-equivalent scenario analysis that includes the combination and indication-specific method patent estate when forecasting Biosimilar entry risk.
Section 7: AI-Assisted Drug Discovery and the Inventorship Crisis
H2: What Happens When the Inventor Is a Machine
Artificial intelligence is deployed throughout modern drug discovery. AlphaFold2 has predicted the 3D structure of nearly every known protein with accuracy that previously required years of crystallography. Generative AI models are designing novel drug candidates de novo by learning from the chemical and pharmacological properties of known drugs. Machine learning classifiers are identifying biomarker signatures in genomic datasets at a scale no human team could replicate. The scientific productivity benefits are substantial and real.
The patent law problem is equally real and largely unresolved. Every major jurisdiction, including the United States, the European Union, and China, requires that an inventor be a natural person. An AI system cannot be named as an inventor. This was confirmed explicitly in the United States in Thaler v. Vidal (Fed. Cir. 2022), where the Federal Circuit upheld the USPTO’s rejection of applications naming DABUS, an AI system, as the sole inventor. The same result has been reached in the United Kingdom, Australia (initially allowing then reversing), and Germany.
H3: The Documentation Imperative
The human inventorship requirement is not a technicality. It imposes a substantive obligation: when AI systems materially contribute to a discovery, the company must be able to identify which human scientists made ‘significant contributions’ to the conception of the claimed invention. This is the same legal standard applied in joint inventorship disputes, and it is a higher bar than many R&D organizations currently meet.
‘Significant contribution’ to conception requires more than owning or operating the AI system. It requires that a human inventor did one or more of the following: framed the specific scientific problem in a way that shaped the AI’s output, designed or selected the training data with scientific judgment about what the model needed to learn, interpreted the AI’s output using scientific expertise to identify which results were genuinely novel and which were artifacts, or made creative choices about how to reduce the AI’s output to a specific working invention.
If the human role was limited to running a standard pipeline on a public dataset and accepting whatever the AI produced, the human contribution to conception may be insufficient to constitute inventorship, even though a human is needed on the application. This creates a risk that competitors could challenge the patent on the grounds that the named inventors did not actually invent the claimed subject matter.
The practical response is a systematic upgrade in R&D documentation protocols. Electronic lab notebooks used in AI-assisted drug discovery programs should record: the scientific question posed to the AI system and the reasoning behind its formulation, the specific model architecture or pipeline used and why that architecture was chosen, the process by which human scientists evaluated, filtered, and selected among the AI’s candidate outputs, and the subsequent experimental validation steps that confirmed the AI-generated hypothesis in vitro or in vivo. This record is the evidence that will support (or undermine) inventorship claims in future litigation.
H3: AI and the Obviousness Problem
AI is not only creating inventorship documentation challenges. It is raising the non-obviousness bar for all biomarker and drug discovery patents, not just those generated by AI systems.
When a patent examiner or PTAB judge evaluates the non-obviousness of a claim, they ask what a person of ordinary skill in the art would have found obvious at the time of filing. If a person of ordinary skill is now expected to have access to and facility with AI analysis tools, including the ability to train models on public genomic databases, then discoveries that emerge from such analyses may be characterized as predictable, not inventive.
This is not merely theoretical. PTAB has begun citing the availability of machine learning tools as a factor in obviousness analyses in some pharmaceutical cases. The implication is that a claim on a biomarker discovered by training a standard classifier on a publicly available dataset may face a stronger obviousness challenge than the same discovery made in 2010 without computational tools. The ‘unexpected results’ safe harbor becomes more important, not less, as AI tools lower the cost of computational biomarker discovery.
H3: Patenting the Algorithm vs. Patenting the Output
When an AI-assisted discovery generates a novel drug candidate or biomarker, a company faces a choice between two patent strategies: patent the discovery itself (the drug molecule or the biomarker) or patent the AI system and method that generated it.
Patenting the discovery is the natural first choice, but as discussed above, it faces eligibility challenges under Mayo if the discovery is a natural correlation, documentation challenges around human inventorship, and non-obviousness challenges if the discovery method is seen as straightforward.
Patenting the AI system has different but complementary advantages. A novel machine learning architecture trained specifically for drug-target interaction prediction, a proprietary training dataset with unique curation methodologies, or a novel data processing pipeline for multi-omics biomarker identification can each be claimed as software-implemented inventions. Under the Alice framework, software claims must be directed to a ‘specific improvement’ in computer functionality or to a non-abstract technological improvement. Claims that recite a specific neural network architecture designed to solve a defined technical problem in bioinformatics, rather than just implementing a generic algorithm, have the best prospect of surviving Alice scrutiny.
The strategic value of algorithm patents extends beyond protecting a single discovery. A proprietary AI platform that can generate drug candidates or biomarker signatures across multiple therapeutic areas is a platform asset with diversified revenue potential. Algorithm patents on the platform create IP protection that is independent of any single drug output, providing value throughout the platform’s productive life.
Key Takeaways: Section 7
The human inventorship requirement mandates that AI-assisted drug discovery programs generate and retain contemporaneous documentation of the specific scientific contributions human inventors made to each claimed conception. This is a compliance and litigation preparedness requirement, not a formality. AI is raising the non-obviousness bar for biomarker patents across the board; unexpected results arguments must be quantitative and specific. Companies should evaluate parallel patent strategies covering both the AI-generated discovery and the AI platform that generated it, because platform patents have diversified value that is not dependent on the commercial success of any single output.
Section 8: CRISPR Gene Editing — The Patent Thicket and How to Operate Within It
H2: The Most Complex Freedom-to-Operate Problem in Modern Biotechnology
CRISPR-Cas9 gene editing is the most powerful personalized medicine technology developed in the past decade. It enables, in principle, the precise correction of any defined genetic mutation in any accessible cell type. For monogenic diseases like sickle cell anemia, beta-thalassemia, Duchenne muscular dystrophy, and Huntington’s disease, it offers a path to curative therapy rather than chronic management. The FDA approvals of Casgevy (exagamglogene autotemcel, developed by Vertex and CRISPR Therapeutics) and Lyfgenia (lovotibeglogene autotemcel, developed by bluebird bio) for sickle cell disease and beta-thalassemia in December 2023 confirmed that CRISPR-based therapies work in humans at clinically meaningful scale.
The IP landscape surrounding CRISPR is the most complex in biotechnology. The foundational patents covering CRISPR-Cas9 use in eukaryotic cells, including human cells, are distributed across multiple institutional holders following a years-long interference and inter partes review proceeding between the Broad Institute of MIT and Harvard, and the University of California, Berkeley, with additional claims from other academic institutions and companies. The current state is not a clear victory for either side but an intricate web of overlapping rights, licensing agreements, sublicensing arrangements, and territorial restrictions.
H3: The Broad-Berkeley Dispute — Current Patent Landscape
The Broad Institute secured patents covering CRISPR-Cas9 genome editing in eukaryotic cells, claiming priority to a 2012 filing by Feng Zhang. The University of California holds patents on the CRISPR-Cas9 system based on the 2012 work of Jennifer Doudna and Emmanuelle Charpentier, which was performed in vitro and demonstrated the basic biochemical mechanism.
The Patent Trial and Appeal Board and the Federal Circuit have ruled in separate proceedings on whether these patents claim patentably distinct inventions (Broad’s position) or whether the eukaryotic cell work was an obvious extension of the in vitro biochemistry (Berkeley’s position). The Broad’s eukaryotic cell claims have survived interference proceedings, and the Federal Circuit affirmed the separation in 2022, preserving the Broad’s U.S. patent estate for CRISPR-Cas9 in human cells.
For a therapeutic developer, the practical consequence is that a license from the Broad is generally required for CRISPR-Cas9-based human cell therapies in the United States. Licenses from Berkeley or its sub-licensees may be required in other jurisdictions or for applications that do not fall within the Broad’s specific claims. Editas Medicine, Intellia Therapeutics, and CRISPR Therapeutics each hold different subsets of licenses from different institutional holders with different field-of-use restrictions.
A developer without the appropriate licenses faces potential injunctive relief that could block clinical trial initiation or commercialization, royalty demands that can substantially reduce program economics, and litigation costs that, in the CRISPR space, have reached nine figures. Freedom-to-operate analysis must be performed before entering any CRISPR-based therapeutic program, and the analysis must cover at minimum the Broad, Berkeley, UC, and Harvard patent estates across all target jurisdictions.
H3: IP Opportunities Beyond the Foundational Patents
The foundational patent thicket covers the core Cas9 enzyme and its use to edit eukaryotic cells. The innovation space beyond the foundational patents is substantial and less crowded.
Novel guide RNA sequences targeting specific disease-relevant mutations are patentable as compositions of matter when the specific guide RNA structure is non-obvious given the prior art. The therapeutic utility of a guide RNA depends heavily on off-target editing rates, and guide RNAs with documented low off-target profiles in a specific gene target are genuinely novel and non-obvious compared to guide RNAs designed by standard algorithmic tools.
Delivery system innovation is a major IP opportunity. Getting the CRISPR machinery into the correct cell type in vivo requires delivery vehicles whose properties, cargo loading efficiency, cell type tropism, and immunogenicity profile are engineering challenges that generate patentable compositions. Lipid nanoparticle (LNP) formulations optimized for hepatocyte delivery (the target for many metabolic disease applications), adeno-associated virus (AAV) variants engineered for specific tissue tropism, and non-viral delivery systems designed for hematopoietic stem cell editing each represent discrete IP opportunities independent of the foundational Cas9 patents.
Next-generation editing tools, including base editors (which make single-nucleotide changes without creating double-strand breaks) and prime editors (which can make precise insertions, deletions, and any type of base conversion), are covered by separate patent estates held primarily by David Liu’s laboratory at the Broad Institute. These tools have distinct patent landscapes from the original Cas9 system and may offer lower royalty burdens for specific applications depending on the field of use.
CRISPR-based diagnostics, using Cas13 or Cas12 systems for highly sensitive nucleic acid detection, represent a fourth IP opportunity category. SHERLOCK (from Broad/MIT) and DETECTR (from Mammoth Biosciences/Berkeley) are competing CRISPR diagnostic platforms with distinct patent estates. These diagnostic applications are outside the scope of the foundational therapeutic CRISPR patents and offer an entry point into CRISPR IP without the full thicket exposure of the therapeutic development programs.
H3: Licensing Strategy and the Gatekeeper Model
Because most therapeutic developers cannot clear the foundational CRISPR thicket independently, a class of licensing intermediaries has emerged. These organizations hold broad licenses from multiple foundational patent holders and sublicense them to therapeutic developers, often with field-of-use restrictions (specific disease indications or genetic targets), territorial restrictions, and milestone and royalty obligations.
The gatekeeper model simplifies access but adds cost and constrains strategic flexibility. A company that obtains a sublicense for sickle cell disease CRISPR therapy, for example, may find that its license does not extend to related hemoglobinopathies, or that the royalty stack from multiple sublicenses materially reduces program economics.
The strategic response is to acquire foundational IP of your own. Companies that hold patents on delivery systems, guide RNA chemistry, or manufacturing processes for cell-based CRISPR therapies gain cross-licensing leverage. Instead of paying one-sided royalties to the foundational patent holders, they can negotiate cross-licenses that reduce their net royalty burden in exchange for access to their own IP. Building a CRISPR IP position strong enough to use as cross-licensing currency is a multi-year, multi-filing effort, but it is the only structural defense against the gatekeeper model’s long-term cost escalation.
Key Takeaways: Section 8
CRISPR-Cas9 foundational patents are concentrated at the Broad Institute and UC Berkeley, with the Broad holding strong rights in eukaryotic cells in the United States. Any therapeutic developer must conduct thorough freedom-to-operate analysis covering all institutional holders before entering a CRISPR program. IP opportunities beyond the foundational patents exist in guide RNA composition, delivery system formulation, next-generation editing tool applications, and CRISPR-based diagnostics. Building a proprietary CRISPR IP portfolio in ancillary technologies creates cross-licensing leverage that reduces the royalty burden imposed by the foundational patent thicket.
Section 9: Global Patent Strategy — U.S., EU, and China
H2: Why a Single Global Filing Strategy Does Not Work
The United States, the European Union, and China represent the three largest pharmaceutical markets and have materially different patent laws, regulatory frameworks, and enforcement environments for personalized medicine innovations. A patent application that claims a diagnostic method in straightforward terms may be granted in one jurisdiction, rejected on formal grounds in another, and granted in a modified form in a third. Companies that draft one application and file it globally without jurisdiction-specific adaptation leave significant protection gaps.
H3: United States — Post-Mayo Prosecution Tactics
The U.S. post-Mayo environment requires diagnostic method claims to include an ‘inventive concept’ beyond a natural law. The most reliable prosecution tactics for personalized medicine claims in the U.S. are:
Claim the specific technical method of detection, not the clinical inference. A claim reciting a specific electrochemical detection method, a proprietary PCR primer set with non-obvious hybridization characteristics, or a defined algorithm for scoring multi-analyte results is more likely to pass Mayo Step 2 than a claim that says ‘detecting the level of biomarker X and adjusting treatment accordingly.’
Use ‘wherein’ clauses to specify non-conventional steps. If the detection method was not commercially available at the time of filing and required inventive work to develop, document this explicitly in the specification and use it in prosecution to distinguish conventional methods.
Draft parallel method-of-treatment claims that tie the diagnostic result to a specific treatment protocol. The Vanda pathway gives method-of-treatment claims in genotype-defined populations a better eligibility posture than pure diagnostic method claims.
File continuation applications as clinical data accumulates, capturing new method-of-treatment claims in newly identified sub-populations with fresh priority dates. This builds out the method claim ring of the IP web over time.
The USPTO’s 2019 Revised Guidance on patent subject matter eligibility provided some clarification on Step 2A of the Alice-Mayo framework, but its practical impact on diagnostic claim prosecution has been modest. Examiners continue to apply Section 101 rejections aggressively to diagnostic method claims. Plan for a prosecution history that includes at least two rounds of Section 101 arguments, with examiner interviews to establish a clear record of the inventive concept.
H3: European Patent Office — ‘For Use’ Claim Strategy
The European Patent Convention explicitly excludes methods for treatment and diagnostic methods practiced on the human body from patentability under Article 53(c). This sounds more restrictive than the U.S. framework, but in practice the EPO’s pragmatic workarounds often produce more predictable outcomes.
The primary workaround is the Swiss-type claim or its successor, the ‘purpose-limited product claim’ (now preferred post-G 2/08): ‘Compound X for use in treating disease Y in patients having biomarker Z.’ This format protects the therapeutic use without claiming the method of treatment directly. Novelty is assessed against prior uses of the compound, and a new biomarker-defined patient population provides the basis for a novel ‘for use’ claim.
For diagnostics, the EPO allows patents on the physical diagnostic kit and its components even where the method of diagnosis is excluded. Claiming the specific antibody, oligonucleotide probe, or enzyme used in the diagnostic process, combined with ‘for use in diagnosing condition Y’ language, creates meaningful protection.
The EPO’s clinical-situation novelty doctrine for sub-population claims, discussed in Section 3, provides a further tool. Claims directed to a biomarker-defined sub-population are generally prosecuted more smoothly at the EPO than in the U.S. because the EPO has an established doctrine for their novelty analysis, while U.S. practice is less settled.
The Unified Patent Court, operational since June 2023, changes the European enforcement landscape substantially. A UPC judgment has effect across all EU member states that have ratified the UPC Agreement. This means a single invalidation action can knock out protection across 17 countries in one proceeding, which increases the downside of patent enforcement for rights holders but also reduces the cost and time of asserting rights. Companies must now make an explicit strategic choice about whether to opt their European patents into or out of the UPC, with ‘opt out’ protecting against central revocation and ‘stay in’ enabling single-forum enforcement.
H3: China — Speed, Biomarker Practice, and Government Alignment
China’s patent system has undergone rapid maturation. It surpassed the United States in total patent application filings in 2011 and has maintained that lead. Specialized IP courts in Beijing, Shanghai, and Guangzhou provide faster, more technically sophisticated adjudication than general civil courts. The average time from infringement filing to judgment in an IP court has dropped to under two years in straightforward cases.
For personalized medicine patent prosecution in China, four features define the strategic environment.
China operates a strict first-to-file system with no grace period for publications by the inventor. Filing before any publication, including conference presentations, is mandatory to preserve priority. Many foreign companies have lost Chinese patent rights to their own inventions because a publication preceded the Chinese filing date by even one day.
The China National Intellectual Property Administration (CNIPA) applies a ‘clear industrial applicability’ standard to biomarker and genetic sequence patents. A patent on a gene sequence or biomarker must clearly identify a specific, practical utility: a diagnostic function, a therapeutic target, a manufacturing tool. Abstract statements of potential utility are insufficient. Companies filing personalized medicine patents in China must include strong utility language tied to a specific practical application in the specification.
Methods of diagnosis and treatment are not patentable under Chinese patent law, mirroring the European position. The workaround is similar: claim the diagnostic kit, the specific reagent, the pharmaceutical composition, or the use of the compound in treating a specific patient population. Chinese examiners apply these exclusions somewhat more broadly than the EPO, so claim drafting must be careful to frame claims around the product or composition rather than the method.
Government policy alignment is a competitive factor unique to China. The Chinese government’s Made in China 2025 and Healthy China 2030 national plans specify domestic precision medicine capability as a strategic priority. Chinese patent examiners are explicitly directed to support examination of domestic precision medicine applications. Foreign companies can leverage this alignment by framing their patent applications to emphasize the health system utility of the innovation in the Chinese population, referencing population-specific biomarker prevalence data where available.
H3: Comparative IP Filing Checklist for a Personalized Medicine Invention
When filing a new personalized medicine patent family globally, the prosecution team should address the following jurisdiction-specific requirements:
United States: prepare arguments for Section 101 eligibility before examination begins; ensure the specification distinguishes the detection method from conventional techniques; draft parallel method-of-treatment claims using the Vanda pathway; prepare continuation applications to capture future biomarker sub-population data.
EPO: draft primary claims in ‘for use’ format for the therapeutic; claim the diagnostic kit components as compositions of matter; prepare novelty arguments under the clinical-situation doctrine for sub-population claims; evaluate UPC opt-out strategy based on portfolio breadth and enforcement risk profile.
China: file before any publication; include specific industrial applicability language tied to a defined practical use; frame claims around the product or composition rather than the method; consider referencing Chinese population biomarker prevalence data to strengthen utility arguments.
Key Takeaways: Section 9
U.S. prosecution strategy for personalized medicine claims must focus on non-conventional detection methods and treatment-oriented claims to survive Mayo. EPO prosecution uses ‘for use’ claim formats and kit composition claims to work around the method exclusion. The UPC has centralized European patent enforcement, raising both the potential reward and the potential risk of European patent assertion. China requires pre-publication filing, specific industrial applicability disclosure, and product-oriented claim drafting; government policy alignment with domestic precision medicine goals can be leveraged in prosecution.
Section 10: Ethical Dimensions — Access, Genetic Privacy, and the Trade-Off at the Core of Patent Law
H2: The Structural Tension Between Incentive and Access
Patent law grants exclusivity as an incentive for disclosure and innovation. The theory is that without exclusivity, innovators would not invest in costly, risky development, and new therapies would not be created. The empirical support for this theory in pharmaceutical development is strong: the industries with the highest R&D investment as a percentage of revenue are those with strong patent protection. The biotechnology sector invests roughly 20 to 25 percent of revenue in R&D, a figure that has no analogue in industries with weak IP protection.
The access problem is equally real. Personalized cancer therapies commonly cost between USD 100,000 and USD 500,000 per patient per year in the United States. Gene therapies, including the approved CRISPR-based sickle cell therapies Casgevy and Lyfgenia, are priced at USD 2.2 million and USD 3.1 million per patient, respectively. These prices are defensible on a cost-effectiveness basis when the therapy is curative and the disease is severe, but they create access barriers that are material and disproportionate by socioeconomic status and insurance coverage.
This is not a problem that patent strategy can solve or that patent teams should be expected to solve alone. But it is a problem that patent strategy affects, and companies that actively build policies to address it, through tiered pricing for emerging markets, voluntary licensing for specific geographic regions or patient populations, and early engagement with government payers on access agreements, reduce the political and regulatory risk that excessive pricing generates. The Novartis approach to Kymriah (tisagenlecleucel), where the company offered outcome-based contracts that refunded the cost of therapy for patients who did not respond at 30 days, illustrates how access and IP exclusivity can be reconciled with creative contracting.
H3: Genetic Data Privacy — GINA’s Limitations and the Data Governance Gap
Personalized medicine generates and processes some of the most sensitive personal data in existence: the complete genetic sequence of an individual patient, often including information about disease predispositions, ancestry, and family health risks. The Genetic Information Nondiscrimination Act of 2008 prohibits U.S. health insurers and employers from using genetic information in coverage or employment decisions. It does not cover life insurance, disability insurance, or long-term care insurance.
This is a material gap in the regulatory framework. A patient who undergoes whole-genome sequencing as part of a personalized medicine program may be generating information that is legally accessible to life and disability insurers making underwriting decisions. This creates a rational incentive to decline genetic testing for patients who would benefit clinically, which directly undermines the personalized medicine model.
For companies in the personalized medicine space, this matters for two reasons. Clinical trial recruitment depends on patients being willing to undergo genomic testing. If patients fear downstream insurance consequences, recruitment for genomic-stratified trials will be harder and more expensive. Post-approval diagnostic adoption depends on the same willingness. Any policy development or corporate advocacy that expands GINA’s scope to cover life and disability insurance directly reduces this barrier to their own commercial success.
The data governance question around biobanks and large genomic databases is separate and equally important. The research databases that underpin biomarker discovery, including the UK Biobank, the All of Us Research Program, and proprietary databases at companies like Regeneron (the GHS database built in partnership with Geisinger Health System), are IP assets with value that extends beyond any single patent. Proprietary access to a well-curated genomic database with linked clinical outcome data is a structural competitive advantage in personalized medicine research. Conversely, the terms under which patient data was collected, the consent framework governing its reuse for commercial purposes, and the regulatory treatment of database-derived discoveries are all sources of legal exposure that require ongoing governance investment.
H3: The Myriad Legacy — Access Lessons From the BRCA Patent Dispute
Myriad Genetics held patents on the isolated BRCA1 and BRCA2 genes and enforced them vigorously, making it the sole U.S. provider of BRCA testing at several thousand dollars per test. The practical consequence was that patients who received an ambiguous initial result could not seek a second opinion from another laboratory, because any lab that performed the test would infringe the patents. Patients who could not afford the test or whose insurance declined coverage had no alternative.
The plaintiffs in the Supreme Court case, including the American Civil Liberties Union, patient advocacy groups, and individual patients, argued that the patents restricted access to life-saving diagnostic information. After the Court’s 2013 ruling invalidated the natural gene sequence claims, the price of BRCA testing dropped substantially within months, and multiple competing test providers entered the market.
The Myriad case is a specific data point about what happens when IP exclusivity in diagnostics reaches its logical extreme without access considerations. For companies with dominant CDx IP positions, the lesson is that monopolistic enforcement of diagnostic patents will eventually generate political and regulatory pressure that is costly to resist, even if legally defensible. Voluntary licensing programs, tiered pricing, or partnership agreements with reference labs in underserved markets reduce this risk at relatively low cost compared to the political capital spent defending an absolute exclusivity position.
Key Takeaways: Section 10
The tension between patent exclusivity as innovation incentive and access to care as a health equity obligation is structural and permanent. Companies that build access frameworks proactively, through outcome-based contracts, tiered pricing, or voluntary licensing, reduce their political and regulatory risk while advancing access goals. GINA’s coverage gaps create a measurable barrier to clinical trial recruitment and diagnostic adoption that is addressable through policy engagement. The Myriad case demonstrates that diagnostic monopolies without access provisions generate litigation, political pressure, and regulatory intervention that erodes the IP position more rapidly than voluntary access programs would have.
Section 11: Patent Intelligence as a Competitive Weapon
H2: What Patent Data Reveals That No Other Source Does
A pharmaceutical company’s patent filing activity is the most reliable public signal of its R&D intentions available. Unlike earnings calls, which are subject to strategic messaging, or scientific publications, which have a 12 to 24 month lag from discovery, patent applications are filed before clinical programs are announced and typically reveal R&D focus years ahead of market awareness.
Systematic patent landscape analysis, conducted at regular intervals across competitor portfolios, reveals competitive entry timing, emerging technology investments, ‘white space’ opportunities in underprotected disease areas, and freedom-to-operate risk before it becomes litigation risk. It does this before any other information source.
H3: What to Look for in a Competitor Patent Landscape
A useful competitive patent landscape analysis examines four categories of information.
First, filing volume and trajectory. A competitor that filed 10 personalized medicine patents per year from 2018 to 2021 and has filed 35 per year since 2022 is telling you something about its strategic priorities without saying a word in public. The acceleration is the signal.
Second, technology category distribution. Comparing the technology category distribution (composition vs. method vs. software vs. formulation) across competitor portfolios reveals their IP architecture philosophy. A company filing primarily method claims in a post-Mayo environment either has an unusually strong position on inventive concept arguments or is building an estate with systemic eligibility risk.
Third, target and biomarker overlap. When multiple competitors are filing patents around the same biological target or the same biomarker, you are seeing a developing competitive cluster. Getting there first, with broader claims and earlier priority dates, matters significantly in these clusters.
Fourth, CDx and biomarker filing timing relative to therapeutic program milestones. A company that files CDx-related patents at Phase II rather than Phase III is demonstrating more sophisticated CDx IP discipline. Compare your own CDx filing timeline against competitors as a benchmark.
H3: Patent Databases — Capability Comparison
The quality of a competitive patent landscape analysis depends on the data infrastructure supporting it. Standard patent databases, including the USPTO’s Patent Full-Text Database, Espacenet, and Google Patents, provide searchable access to issued patents and published applications. They do not provide pharmaceutical-specific curation, orange book linkage, regulatory approval status, litigation history integration, or expiry date calculation adjusted for patent term extension.
Specialized pharmaceutical patent intelligence platforms like DrugPatentWatch combine patent data with FDA Orange Book listings, regulatory approval histories, and paragraph IV filing records to build the cross-referenced picture that pharmaceutical IP teams actually need. For a specific drug patent, DrugPatentWatch aggregates: composition patents, formulation patents, method of use patents, all Hatch-Waxman paragraph IV filings against those patents, litigation outcomes, and projected generic entry dates. For CDx patents, it provides linkage to the drug’s regulatory status and labeling history.
For generic and biosimilar manufacturers, this infrastructure identifies patent expiration windows across the portfolio before making development investment decisions. For branded companies, it identifies which of their patents are actually listed in the Orange Book, which have been challenged, and where there are unchallenged gaps in coverage that competitors could exploit. Neither use case can be served adequately by a general-purpose patent database.
H3: ‘White Space’ Identification for R&D Priority Setting
A white space in a patent landscape is a technology or indication area with meaningful unmet medical need and relatively low existing patent coverage. White spaces represent opportunities to enter a new area with a reasonable prospect of claiming broad protection, because the foundational IP is not yet established.
Identifying white spaces requires a two-step analysis: mapping existing patent density against disease areas with high unmet need, and then assessing whether the low patent density reflects early-stage opportunity or intractable scientific difficulty. Some white spaces are empty because no one has found a viable approach to the problem. Others are empty because the research has not yet scaled sufficiently to generate patents, even though the science is promising.
For personalized medicine, current white spaces include: multi-biomarker combination diagnostics for treatment selection in diseases with heterogeneous molecular drivers (e.g., Alzheimer’s disease, treatment-resistant depression); pharmacogenomic dosing algorithms for commonly prescribed drugs where the CYP450 interaction data exists but has not been clinically validated to patent-defensible levels; and CRISPR-based diagnostics using novel Cas effector proteins with different cleavage specificities than Cas9 or Cas12.
Key Takeaways: Section 11
Patent data is the most forward-looking competitive intelligence available, revealing R&D intent years before public announcement. Competitive patent landscape analysis should track filing volume trajectory, technology category distribution, biomarker and target overlap, and CDx filing timing relative to therapeutic milestones. Pharmaceutical-specific patent intelligence platforms, which integrate patent data with Orange Book listings, regulatory approvals, and paragraph IV filing records, provide materially more actionable analysis than general-purpose patent databases. White space identification, combining low patent density with high unmet medical need, is a primary tool for prioritizing R&D investment in areas with near-term opportunity for broad protection.
Comprehensive Investment Strategy: Personalized Medicine IP for Portfolio Managers
H2: How to Value a Personalized Medicine IP Estate
Personalized medicine IP valuation requires moving beyond simple patent counting and expiry-date analysis. The most common valuation error in pharma equities is equating primary composition-of-matter patent expiry with franchise revenue cliff. The Herceptin and Keytruda case studies demonstrate that this error systematically undervalues companies with active CDx innovation programs and broad method-of-treatment claim estates.
A more rigorous framework values the IP estate in layers. The primary composition-of-matter claims generate the core exclusivity period and carry the highest per-patent revenue attribution. The CDx patent cluster generates indirect revenue through therapeutic enablement and direct revenue through diagnostic sales; its value should be assessed based on the patient selection gateway function, not just kit revenue. The method-of-treatment patent ring, covering specific patient sub-populations and dosing regimens, extends exclusivity at lower individual patent value but with high aggregate value because it constrains biosimilar commercial use even after the composition patent expires. The combination therapy patent ring extends exclusivity into the era of next-line treatment and is often underweighted in sell-side analysis.
For early-stage biotechs, IP estate quality is a primary predictor of long-term value capture. Two companies with equal Phase II data and equal capital can diverge dramatically in long-term valuation based on whether their IP architect built a multi-layer web or filed a handful of composition claims without CDx or method-of-treatment coverage. Due diligence on IP quality should be conducted by specialized pharmaceutical IP counsel before Series B financing and certainly before IPO. Summary representations in prospectuses are not a substitute for independent claim-by-claim analysis.
For established pharmaceutical companies, the most informative metric is not total patent count but ‘post-primary-expiry coverage,’ defined as the percentage of current revenue that will remain patent-protected after the primary composition-of-matter patent expires. A company with USD 5 billion in annual drug revenue and 80 percent post-primary coverage (through CDx, formulation, and combination patents) has a materially more durable franchise than a company with USD 10 billion in revenue and 20 percent post-primary coverage.
H3: M&A Due Diligence Checklist for Personalized Medicine Targets
Any acquisition of a personalized medicine asset should include at minimum:
An independent Section 101 eligibility audit of all diagnostic method claims in the portfolio, scored by a qualified patent litigator with Federal Circuit diagnostic claim experience, not outside counsel who prosecuted the patents.
A freedom-to-operate analysis for the companion diagnostic against the foundational CDx patent estates of Dako/Agilent, Roche, Abbott/Alinity, and any disease-specific CDx patent holders in the relevant biomarker space.
A paragraph IV history review identifying which of the target’s patents have been challenged, with litigation outcomes and any claim amendments or disclaimers made during litigation that could narrow the protection scope.
A comparison of CDx filing dates against therapeutic development milestones to assess whether the CDx IP was developed in parallel with the therapeutic program or as an afterthought. Late-stage CDx patent filing is a red flag for thin protection on the patient selection gateway.
An inventory of trade secrets, including any LDT-based diagnostic algorithms not covered by patents, with assessment of their actual confidentiality status. Trade secrets that have been described in publications, grant applications, or conference presentations are not trade secrets.
H3: Key Metrics for Personalized Medicine IP Portfolio Benchmarking
Analysts and IP teams should track the following metrics for benchmarking purposes across peer companies:
Patent filing rate per research scientist (annual applications per 100 FTEs in R&D): a measure of IP culture and disclosure rate.
Ratio of composition-of-matter claims to method claims in the total portfolio: higher composition ratios indicate stronger foundational protection; high method claim ratios in a post-Mayo environment indicate eligibility risk.
Percentage of CDx patents filed at Phase II or earlier versus Phase III or later: earlier filing reflects stronger CDx IP discipline.
Post-primary-expiry coverage ratio as defined above: the single most informative measure of franchise durability.
PTAB challenge rate against the portfolio: the percentage of the company’s patents that have been subjected to inter partes review or post-grant review petitions, and the percentage of challenged claims that survived. High challenge rates combined with low survival rates signal systematic claim drafting weaknesses.
Frequently Asked Questions
Q: After Mayo, is there any reliable strategy for patenting a diagnostic method that relies on a biomarker correlation?
The most reliable strategy combines three elements. First, claim the specific technical method of detecting the biomarker, using language that distinguishes the claimed detection method from conventional techniques available at the time of filing. Second, draft parallel method-of-treatment claims that tie the biomarker result to a specific treatment decision, using the Vanda pathway. Third, claim the physical diagnostic kit and its specific reagent components as compositions of matter, which are not subject to Mayo. No single claim type is fully reliable post-Mayo, but this combination provides multiple avenues to enforce the IP even if one avenue is blocked.
Q: Our company uses AI models extensively in drug discovery. What do we need to do differently in the lab to protect IP?
Implement a specific documentation protocol for AI-assisted discovery programs that captures: the scientific problem framing done by the human team before engaging the AI, the rationale for the training data curation and model selection, the process by which human scientists evaluated and selected among the AI’s candidate outputs using scientific judgment, and the experimental validation steps that followed. This documentation is evidence that human inventors made ‘significant contributions’ to the conception of each claimed invention. Standard electronic lab notebooks are usually adequate if configured to capture these specific decision points. Do this before filing, not after an inventorship challenge arises.
Q: Our lead drug’s companion diagnostic was approved five years ago and newer competitors have better diagnostic performance. How should we respond?
Treat CDx performance obsolescence the same way a branded drug company treats a competitor’s superior efficacy data: respond with your own. Commission head-to-head clinical validation studies that compare your CDx to the competitor’s on the metrics that matter to oncologists and payers (sensitivity, specificity, concordance with genomic sequencing, turnaround time, and cost). If your CDx underperforms, consider licensing or acquiring the superior technology and pursuing a label update rather than defending an inferior product. The regulatory lock-in that protects your approved CDx is not permanent: if payers start requiring a competitor’s test as a condition of reimbursement approval, the commercial protection erodes even without a formal label change.
Q: We are developing a CRISPR-based therapy. Where do we start on freedom-to-operate?
Start with the four institutional patent estates most likely to have relevant claims: Broad Institute, UC Berkeley, UC San Francisco, and the Whitehead Institute. Each has a technology licensing office that will discuss freedom-to-operate and licensing terms. Supplement this with a full-text patent database search across CNIPA, EPO, and USPTO covering all Cas9, Cas12, and Cas13 variants, all guide RNA synthesis methods, and all eukaryotic cell editing method claims. Most importantly, engage outside counsel with specific CRISPR patent experience — this is not a patent landscape that should be assessed by counsel who have not been through the Broad-Berkeley interference proceedings.
Q: How should we structure CDx IP ownership in a co-development agreement with a diagnostic partner?
The agreement should specify IP ownership of the approved CDx separately from improvements made post-approval. The most common model gives the diagnostic company ownership of the CDx IP and grants the drug company an exclusive license to use the CDx for the specific labeled indication. Improvements to the CDx that are generated during the co-development relationship should have a defined ownership rule — either joint ownership with defined licensing rights, or automatic assignment to the generating party with a cross-license. Define what happens to the CDx IP if the drug is acquired: does the acquirer inherit the exclusive CDx license? What are the exclusivity scope limitations? Address these questions in the initial agreement. Revisiting them during an acquisition negotiation is significantly more expensive.
Final Observations
The personalized medicine patent landscape in 2025 is simultaneously more constrained and more opportunity-rich than it was a decade ago. More constrained, because Mayo has eliminated a category of diagnostic method claims that previously generated substantial patent value, because the CRISPR thicket has made freedom-to-operate analysis a prerequisite for gene therapy entry, and because AI-assisted discovery is raising the non-obviousness bar across all biomarker research.
More opportunity-rich, because the CDx market is growing at 15 percent annually and remains underprotected relative to its commercial importance, because CRISPR delivery and guide RNA chemistry offer IP opportunities outside the crowded foundational thicket, because AI platform patents can protect the tools of discovery independently of any individual discovered compound, and because the combination therapy and biomarker expansion strategies available to precision oncology companies provide an evergreening toolkit that small molecule pharma never had.
Companies that navigate this landscape well will be those that integrate IP strategy into R&D planning at Phase I, not Phase III; that build CDx IP with the same rigor applied to therapeutic IP; that document AI-assisted discovery processes with litigation preparedness in mind; and that approach global prosecution as a multi-jurisdiction optimization problem rather than a PCT filing exercise. The ones that do not will find that their most valuable scientific discoveries are either unpatentable or inefficiently protected — a costly outcome in a market where the prize is measured in hundreds of billions of dollars.
This analysis was prepared as a strategic reference document for pharmaceutical IP teams, R&D leads, and institutional investors. It does not constitute legal advice. Patent eligibility and prosecution strategies should be developed with qualified patent counsel with jurisdiction-specific experience.
Patent data sourced from USPTO, EPO, and CNIPA. Market size data sourced from Precedence Research, ResearchAndMarkets, and BCC Research. Clinical and regulatory data sourced from FDA.gov, published KEYNOTE and other clinical trial records, and peer-reviewed oncology literature.
