{"id":34528,"date":"2025-10-19T14:28:28","date_gmt":"2025-10-19T18:28:28","guid":{"rendered":"https:\/\/www.drugpatentwatch.com\/blog\/?p=34528"},"modified":"2026-04-20T10:16:57","modified_gmt":"2026-04-20T14:16:57","slug":"the-innovation-compass-using-drug-patent-citation-network-analysis-to-chart-the-future-of-pharmaceutical-research","status":"publish","type":"post","link":"https:\/\/www.drugpatentwatch.com\/blog\/the-innovation-compass-using-drug-patent-citation-network-analysis-to-chart-the-future-of-pharmaceutical-research\/","title":{"rendered":"Drug Patent Citation Network Analysis: The Definitive Guide to Predicting Pharma’s Next Moves"},"content":{"rendered":"\n

Why Standard Patent Intelligence Is Failing You<\/strong><\/h2>\n\n\n\n
\"\"<\/figure>\n\n\n\n

The patent cliff is no longer the right mental model. For two decades, pharma strategists framed IP risk as a countdown: one date, one molecule, one revenue drop. That framing is now actively dangerous. The industry’s IP environment today consists of layered patent thickets on biologics, Paragraph IV litigation running in parallel across multiple jurisdictions, platform technology patents that cut across therapeutic areas, and AI-assisted design-around strategies that can appear within months of a filing. Looking at a single expiry date in this environment is the equivalent of navigating a highway by watching the rearview mirror.<\/p>\n\n\n\n

The companies that will control the next decade of pharmaceutical revenue are those that can read the patent landscape as a dynamic, interconnected network rather than a collection of individual documents. Patent citation network analysis provides that capability. It treats patents as nodes and citations as directional edges, producing a topological map of the entire innovation ecosystem. From that map, you can identify foundational ‘crown jewel’ assets, detect emerging research fronts before they reach Phase I, decode a competitor’s full defensive architecture, and locate the uncontested white space where first-mover IP protection is still available.<\/p>\n\n\n\n

This guide is designed for IP teams, portfolio managers, R&D leads, and institutional investors who need that capability now. It starts with the building blocks of citation data, moves through the full analytical toolkit, and applies the framework to the specific competitive challenges of biologics, oncology, and AI-driven drug discovery. Every section is designed to add decision-relevant information, not background.<\/p>\n\n\n\n

Key Takeaways: Why This Matters Now<\/strong><\/p>\n\n\n\n

    \n
  • The ‘single patent cliff’ model is obsolete. High-value drugs now sit behind layered IP architectures involving dozens to hundreds of secondary patents.<\/li>\n\n\n\n
  • Patent citation networks are leading indicators of R&D direction. They show where money and talent are flowing before press releases and trial registrations confirm it.<\/li>\n\n\n\n
  • Data harmonization \u2014 not visualization software \u2014 is the primary source of competitive advantage in patent analytics. The analyst with cleaner corporate tree data wins.<\/li>\n\n\n\n
  • For institutional investors, citation-weighted patent metrics predict R&D productivity significantly better than raw patent counts, with correlation coefficients reaching 0.75 in foundational economic research.<\/li>\n<\/ul>\n\n\n\n
    \n\n\n\n

    The Anatomy of a Patent Citation<\/strong><\/h2>\n\n\n\n

    A patent citation is a reference within a patent application to prior art: any earlier document considered relevant to the novelty or inventive step of the claimed invention. That prior art can be a previously granted patent, a published application, a peer-reviewed journal article, a conference paper, a textbook, a database entry, or even a prior oral disclosure captured in a citable form. The full collection of these references is the patent’s ‘prior art record.’<\/p>\n\n\n\n

    Citations enter the record through two distinct channels, and distinguishing between them is analytically critical. The applicant’s duty of disclosure, codified in 37 CFR 1.56 in the United States, requires that applicants submit all known material prior art. The standard is ‘materiality’: any information that could affect the patentability of at least one claim. Failure to comply exposes a granted patent to later invalidation for inequitable conduct. Patent examiners at the USPTO and the EPO conduct independent prior art searches and add their own references, often including documents the applicant did not submit.<\/p>\n\n\n\n

    The EPO’s search report goes one step further by categorizing prior art with letter codes. An ‘X’ designation indicates that the cited document alone is sufficient to call a claimed invention’s novelty or inventive step into question. A ‘Y’ designation applies when the document, in combination with one or more other ‘Y’ documents, raises the same concern. For an analyst reviewing a competitor’s application, an examiner-placed ‘X’ citation pointing to your company’s foundational patent is an immediate competitive intelligence signal: the examiner has objectively confirmed technological overlap at the highest level of relevance.<\/p>\n\n\n\n

    Non-patent literature (NPL) citations deserve separate treatment in any rigorous analysis. When a pharmaceutical patent cites a journal article published in Nature Medicine<\/em> or JACS<\/em>, it is documenting the linkage between applied commercial R&D and basic research. Studies by Bhaven Sampat and colleagues have shown that NPL citation rates vary significantly across therapeutic areas and company types, and that citations to government-funded research (particularly NIH-funded papers) are disproportionately common in high-value pharmaceutical patents. This makes NPL citation patterns a useful proxy for a company’s depth of engagement with upstream academic science.<\/p>\n\n\n\n

    \n\n\n\n

    Forward vs. Backward Citations: What Each Direction Tells You<\/strong><\/h2>\n\n\n\n

    Every patent citation has a direction, and that direction determines what strategic question it answers.<\/p>\n\n\n\n

    Backward citations<\/strong> (the references a patent makes to earlier documents) reveal the technological genealogy of an invention. Reading a patent’s backward citations is equivalent to reading its technical bibliography: you can trace the foundational science, identify the platform technologies it builds on, and reconstruct the prior art landscape the inventors were working within. For competitive intelligence, a competitor patent with backward citations concentrated in your company’s IP cluster signals either that they are building on your work or that they are constructing a design-around that acknowledges your foundational position. Either scenario demands a response.<\/p>\n\n\n\n

    A sparse backward citation list is its own signal. When a patent cites little prior art, it either covers genuinely novel ground with few predecessors or the applicant has been selective in disclosure. Distinguishing between these possibilities requires cross-referencing the examiner’s search report: if the examiner adds substantial citations that the applicant omitted, the latter scenario becomes more likely.<\/p>\n\n\n\n

    Forward citations<\/strong> (the subsequent patents that cite a given patent) are the primary quantitative proxy for a patent’s technological impact and commercial relevance. The intuition is straightforward: a foundational invention will be referenced by the researchers who build on it, creating a forward citation trail proportional to its influence. High forward citation counts correlate with higher patent value across multiple empirical studies. The correlation is not perfect, but it is strong enough that citation-weighted patent counts outperform raw counts as predictors of R&D output value in pharmaceutical contexts.<\/p>\n\n\n\n

    Citation velocity<\/strong> refines raw forward citation counts by normalizing for patent age: velocity = total forward citations \/ years since first publication. A patent filed in 2022 with 40 forward citations by 2025 has a velocity of ~13 per year. A patent filed in 2008 with 150 forward citations has a velocity of roughly 9 per year. The 2022 patent may be more immediately relevant to current R&D despite having fewer total citations. Velocity is the metric to use when comparing patents across different filing cohorts and when trying to identify which recent inventions are gaining momentum.<\/p>\n\n\n\n

    Key Takeaways: Citation Directionality<\/strong><\/p>\n\n\n\n

      \n
    • Backward citations map a patent’s scientific DNA and reveal dependence on specific IP clusters.<\/li>\n\n\n\n
    • Forward citations measure technological influence and are the most widely validated proxy for patent value.<\/li>\n\n\n\n
    • Citation velocity corrects for age bias and is essential when comparing patents filed in different years.<\/li>\n\n\n\n
    • Examiner-added forward citations to your portfolio from a competitor’s application are high-confidence signals of technological overlap, sourced from a neutral third party.<\/li>\n<\/ul>\n\n\n\n
      \n\n\n\n

      Why Citation Analysis Hits Harder in Pharma Than Any Other Industry<\/strong><\/h2>\n\n\n\n

      Patent protection is the most important appropriability mechanism across all industries for which it has been studied, but its importance in pharmaceuticals exceeds that in every other sector by a substantial margin. A 2018 CBO analysis and the broader economic literature consistently show that pharma companies are the only industry group that ranks patents as either the first or second most important tool for capturing R&D returns, ahead of trade secrets, lead time, and complementary assets.<\/p>\n\n\n\n

      The reason is structural: a small molecule drug is, by design, chemically reproducible. Once the structure is disclosed, a generic manufacturer needs only chemistry and regulatory expertise to replicate it. The patent on the active pharmaceutical ingredient (API) is often the single barrier between a $10 billion revenue stream and immediate generic competition. This creates an exceptionally direct link between a specific patent’s legal status and a specific product’s commercial value.<\/p>\n\n\n\n

      Three features of the pharmaceutical regulatory landscape amplify the strategic utility of citation analysis beyond what is possible in other industries.<\/p>\n\n\n\n

      The FDA’s Orange Book (formally, Approved Drug Products with Therapeutic Equivalence Evaluations<\/em>) explicitly lists every patent a drug sponsor believes covers an approved NDA. This public, structured database allows analysts to directly connect citation network data to named drug products, commercial revenue figures, and Paragraph IV challenge history. In semiconductors or consumer electronics, the relationship between a patent and a specific revenue stream requires inference. In pharmaceuticals, the Orange Book makes it explicit.<\/p>\n\n\n\n

      The Purple Book provides the equivalent linkage for biologics: it lists reference products, their biosimilar applicants, and the biosimilarity or interchangeability determinations made under the BPCIA. Combining Purple Book data with biologic patent network analysis produces a complete picture of which patents stand between a reference biologic and its first biosimilar competitor.<\/p>\n\n\n\n

      Patent term extensions (PTEs) under the Hatch-Waxman Act, and the equivalent Supplementary Protection Certificates (SPCs) in the EU, extend the effective life of a single qualifying patent by up to five years to compensate for regulatory review time. Tracking which patents in a citation network carry active PTEs or SPCs changes the effective expiry date landscape substantially and is a variable that platforms like DrugPatentWatch track directly.<\/p>\n\n\n\n

      Key Takeaways: The Pharma-Specific Advantage<\/strong><\/p>\n\n\n\n

        \n
      • The Orange Book and Purple Book make the link between patents and commercial products explicit, enabling revenue-level IP valuation that is not possible in most other industries.<\/li>\n\n\n\n
      • Patent term extensions under Hatch-Waxman and SPCs in the EU can shift effective exclusivity dates by up to five years relative to the face date on the grant.<\/li>\n\n\n\n
      • The ‘easy to copy’ economics of small molecules mean that a single API patent is often the only structural barrier to generic entry, making its citation network position a direct input to revenue forecasting.<\/li>\n<\/ul>\n\n\n\n
        \n\n\n\n

        Patent Families: Building a Global Picture of a Single Invention<\/strong><\/h2>\n\n\n\n

        No single patent document captures the full strategic reality of an invention. A company protecting a major drug candidate will file a priority application in its home jurisdiction, then file international applications under the Patent Cooperation Treaty (PCT) and national phase applications in all commercially significant markets. The set of all applications sharing the same priority claim constitutes a patent family.<\/p>\n\n\n\n

        The EPO’s DOCDB simple patent family groups applications that cover identical technical content based on shared priority data. The INPADOC extended patent family applies a broader definition, grouping applications that share any priority document in common and therefore cover related but not necessarily identical subject matter. The distinction matters for analysis: a DOCDB family analysis is precise but may miss related applications that were filed with different priority claims; an INPADOC analysis captures more of the strategic landscape but can introduce noise.<\/p>\n\n\n\n

        From a competitive intelligence standpoint, the geographic distribution of a patent family’s members is a direct read of the applicant’s commercial strategy. Filing national phase applications in China, Japan, Germany, the UK, and Canada requires substantial financial investment. Companies do not make that investment for patents they consider marginal. A filing pattern that covers only the US and EU with no PCT extension signals a narrower commercial ambition than one that includes CNIPA filings and Japanese examination. Similarly, the voluntary abandonment of specific family members during prosecution, visible in publicly available legal status data, indicates that the applicant concluded the protection in that jurisdiction was not worth the maintenance costs.<\/p>\n\n\n\n

        For citation analysis, studying the full family rather than the US grant in isolation provides a substantially richer picture. Different examiners in different jurisdictions will conduct independent prior art searches and add different citations. A complete family-level citation analysis therefore surfaces prior art from a broader range of searching perspectives, reducing the risk of missing critical references that appear in non-US search reports.<\/p>\n\n\n\n

        \n\n\n\n

        The Examiner Signal: Why Not All Citations Are Equal<\/strong><\/h2>\n\n\n\n

        The standard approach to patent citation analysis treats all citations as equivalent data points. A citation is a citation. This is a significant analytical error.<\/p>\n\n\n\n

        When an applicant submits a reference under their duty of disclosure, they are making a judgment call, filtered through their legal interests. They disclose what they believe is material, and the threshold for materiality is often interpreted conservatively. Patent counsel representing a major pharma company has every incentive to cast a wide net on disclosure to avoid later inequitable conduct challenges, but they also have an interest in not spotlighting prior art that could be used to narrow or invalidate their client’s claims. The citations list is therefore a strategic document as well as a legal one.<\/p>\n\n\n\n

        When an examiner adds a reference, the dynamic changes. The examiner has no commercial interest in the outcome. Their job is to find the closest prior art and cite it. An examiner-added citation to a competitor’s patent is an objective, third-party confirmation that the examiner reviewing your application found the competitor’s technology relevant enough to include in the formal record.<\/p>\n\n\n\n

        This distinction creates a two-tier citation classification that should be standard in any rigorous analysis. Tier 1 citations are examiner-added, particularly those carrying an EPO ‘X’ or ‘Y’ relevance code. These are the highest-confidence signals of genuine technological proximity. Tier 2 citations are applicant-submitted. They carry information, but require more careful interpretation because they can reflect strategic over-disclosure as easily as genuine technological relevance.<\/p>\n\n\n\n

        Building this distinction into a network model, by weighting Tier 1 edges more heavily than Tier 2 edges, produces a more accurate map of the technological landscape. It filters out the noise of mass-disclosure and amplifies the signal of examiner-confirmed technological overlap. Most commercial patent intelligence platforms do not yet do this automatically. Analysts who build this weighting into their custom models gain a material edge in the quality of their network maps.<\/p>\n\n\n\n

        \n\n\n\n

        Data Quality: The Non-Negotiable Foundation<\/strong><\/h2>\n\n\n\n

        Raw patent data from the USPTO PAIR system, the EPO’s Open Patent Services API, or WIPO’s PATENTSCOPE is not ready for analysis. Every public patent office is an administrative body built to examine and grant patents, not to serve as a structured data provider. The raw data reflects this: assignee names are inconsistently formatted, subsidiary structures are opaque, continuations and continuation-in-part applications introduce complex family structures that require manual resolution, and legal status updates lag the actual procedural events by weeks or months.<\/p>\n\n\n\n

        The consequences of skipping harmonization are not theoretical. An analyst constructing a competitive landscape for a potential acquisition target that fails to link the target’s patents to its parent company will systematically underestimate the portfolio’s true scope. An R&D team conducting a white space analysis on harmonized data may identify what appears to be an unoccupied technology zone, only to discover post-engagement that a major competitor’s subsidiary has been filing there for three years under a name that did not surface in the raw data query.<\/p>\n\n\n\n

        The Data Harmonization Workflow<\/strong><\/p>\n\n\n\n

        The full workflow for producing analysis-ready data involves five sequential steps:<\/p>\n\n\n\n

        Step 1: Multi-Jurisdictional Data Aggregation.<\/em> Pull patent records from the USPTO, EPO, CNIPA, JPO, WIPO, and national phase offices covering all commercially relevant markets. For each record, collect the full metadata: filing date, publication date, grant date, priority claims, inventor list, assignee list, classification codes (CPC and IPC), legal status history, and the full citation record (both backward and forward, with examiner vs. applicant attribution where available).<\/p>\n\n\n\n

        Step 2: Assignee Normalization.<\/em> This is the most time-intensive and analytically critical step. ‘Pfizer Inc.,’ ‘Pfizer Pharmaceuticals LLC,’ ‘Pfizer Ireland Pharmaceuticals,’ and ‘Agouron Pharmaceuticals Inc.’ all represent, at various points and for various products, the same ultimate corporate owner. A database that does not map these to a single entity will produce a fragmented, misleading portfolio picture. High-quality platforms maintain corporate trees that track parent-subsidiary relationships and update them in response to M&A activity, often integrating with commercial corporate structure databases like Bureau van Dijk’s Orbis.<\/p>\n\n\n\n

        Step 3: Family Deduplication.<\/em> Multiple applications covering the same invention in different jurisdictions should be linked into their family structure so that analyses can be conducted at either the application level or the invention level, depending on the question being asked.<\/p>\n\n\n\n

        Step 4: Text Normalization and NLP Pre-processing.<\/em> Patent claims and abstracts are unstructured text. Preparing them for semantic analysis, topic modeling, or chemical entity recognition requires tokenization, stopword removal, stemming or lemmatization, and, for pharmaceutical patents specifically, normalization of chemical nomenclature (INN names, CAS registry numbers, IUPAC names) and biological sequence identifiers.<\/p>\n\n\n\n

        Step 5: Adjacency Matrix Construction.<\/em> The final output is a directed adjacency matrix where each row and column represents a patent node, and a value of 1 at position (i,j) indicates that patent i cites patent j. This matrix, potentially filtered by time window, technology class, or assignee, is the input for all network analysis algorithms.<\/p>\n\n\n\n

        An important note on the intelligence embedded in the harmonization process itself: when an analyst tracks corporate trees closely enough to catch unreported M&A activity reflected in patent assignment records before it reaches the financial press, the data cleaning process has already generated a market-moving insight.<\/p>\n\n\n\n

        Key Takeaways: Data Quality<\/strong><\/p>\n\n\n\n

          \n
        • Assignee harmonization is the single largest source of analytical error in patent network studies and the single largest source of competitive advantage for teams that do it rigorously.<\/li>\n\n\n\n
        • Multi-jurisdictional coverage is mandatory. Limiting analysis to the USPTO misses a substantial fraction of global innovation activity, particularly from Chinese and Japanese applicants.<\/li>\n\n\n\n
        • Patent assignment records in the USPTO’s assignment database can flag M&A activity weeks before public announcement. Teams monitoring these records in near real-time gain a documented information edge.<\/li>\n<\/ul>\n\n\n\n
          \n\n\n\n

          Building the Network: Adjacency Matrices, Nodes, and Edges<\/strong><\/h2>\n\n\n\n

          Once the data is clean, the network is constructed as a directed graph. The patents are nodes (vertices). Each citation relationship is a directed edge running from the citing patent to the cited patent. The direction matters: it encodes the temporal and conceptual flow of knowledge from foundational prior art to subsequent inventions.<\/p>\n\n\n\n

          In mathematical notation, the adjacency matrix A has elements $a_{ij} = 1$ if patent $i$ cites patent $j$, and $a_{ij} = 0$ otherwise. For most pharmaceutical patent datasets, this matrix is extremely sparse: the number of actual edges is a tiny fraction of the total possible edges ($n^2$ for $n$ patents). This sparseness is a feature, not a flaw. It means that the edges that do exist carry strong signal about genuine technological relationship.<\/p>\n\n\n\n

          Network density<\/strong> is calculated as $D = |E| \/ (|V| \\cdot (|V| – 1))$, where $|E|$ is the number of directed edges and $|V|$ is the number of nodes. A dense sub-network within a sparse overall graph identifies a cluster of patents that cite each other heavily, which in patent terms means a focused, inter-referential community of related inventions. These clusters are the primary analytical output of network construction.<\/p>\n\n\n\n

          Clustering algorithms<\/strong> partition the network into these communities. The Louvain method, which optimizes a modularity score to find the partition that maximizes within-community edge density relative to a random baseline, is widely used for large patent networks. For temporal analysis, the InfoMap algorithm, which treats information flow through the network as a random walk, is often preferred because it is more sensitive to the directionality of citations. The choice of algorithm affects the resulting cluster boundaries and should be validated against domain knowledge in the target technology area.<\/p>\n\n\n\n

          Once clusters are identified, they can be characterized by their most distinctive vocabulary (using TF-ICF, the citation-network equivalent of TF-IDF), their most central patents by each centrality measure, their filing date distribution, and the assignees most heavily represented within them. This characterization is the product that gets consumed by R&D leads and portfolio managers.<\/p>\n\n\n\n

          \n\n\n\n

          Centrality Metrics Decoded: What Each Measure Actually Tells a Strategist<\/strong><\/h2>\n\n\n\n

          Network science offers multiple definitions of ‘importance’ within a graph, and each captures a different type of strategic relevance. Using only one is like assessing a drug candidate using only one biomarker.<\/p>\n\n\n\n

          In-Degree Centrality (Forward Citation Count).<\/strong> The number of patents that cite a given node. In a directed patent citation graph, this is the patent’s forward citation count. High in-degree patents are technological authorities: their inventions have been used as building blocks by subsequent innovators. For pharmaceutical IP teams, in-degree centrality identifies the assets most likely to generate licensing revenue, the assets most likely to be challenged (because they are the most commercially significant), and the assets that define the technological core of a competitor’s position in a therapeutic area.<\/p>\n\n\n\n

          Out-Degree Centrality (Backward Citation Count).<\/strong> The number of prior art documents a patent cites. High out-degree patents synthesize a wide range of prior work. In pharmaceutical research, a patent with high out-degree in a new therapeutic area may indicate a systematic, comprehensive literature review, which can signal either a well-resourced discovery program or an attempt to pre-empt freedom-to-operate challenges by documenting broad awareness of the prior art landscape.<\/p>\n\n\n\n

          Betweenness Centrality.<\/strong> Formally, $B(v) = \\sum_{s \\neq v \\neq t} \\frac{\\sigma_{st}(v)}{\\sigma_{st}}$, where $\\sigma_{st}$ is the total number of shortest paths from node $s$ to node $t$, and $\\sigma_{st}(v)$ is the subset of those paths that pass through node $v$. Patents with high betweenness sit on the knowledge bridges between otherwise disconnected technology clusters. They are the interdisciplinary breakthroughs, the platform technologies that connect molecular biology to computational chemistry, or the delivery system patents that sit between a drug cluster and a device cluster. These are often the most strategically valuable assets in a network because they define new areas of convergence. High betweenness centrality is the first metric to query when hunting for white space opportunities at technological intersections.<\/p>\n\n\n\n

          Eigenvector Centrality.<\/strong> A patent’s eigenvector centrality is high if it is connected to other patents that are themselves highly connected. A single citation from a patent that happens to be the most foundational work in its field contributes more to eigenvector centrality than ten citations from marginal, rarely-cited applications. This metric distinguishes true ‘crown jewels’ \u2014 patents whose importance is confirmed by other important patents \u2014 from patents that are simply popular in a specific narrow niche.<\/p>\n\n\n\n

          PageRank.<\/strong> Google’s original web-ranking algorithm applies directly to patent citation graphs. It models the importance of a patent as the probability that a researcher following citation links at random would land on that patent. PageRank handles directed edges correctly and is robust to network irregularities, making it a reliable metric for identifying the most influential documents in large, heterogeneous patent datasets.<\/p>\n\n\n\n

          Table 1: Centrality Metrics and Their Strategic Applications<\/strong><\/p>\n\n\n\n

          Metric<\/th>Technical Definition<\/th>Strategic Meaning<\/th>Primary Use Case<\/th><\/tr><\/thead>
          In-Degree<\/td>Count of incoming citation edges<\/td>Technological authority and influence<\/td>Identify licensing targets; measure competitor IP impact<\/td><\/tr>
          Out-Degree<\/td>Count of outgoing citation edges<\/td>Breadth of knowledge synthesis<\/td>Flag interdisciplinary or comprehensive filings<\/td><\/tr>
          Betweenness<\/td>Frequency on shortest inter-node paths<\/td>Bridge position between technology clusters<\/td>Find convergence white space; locate interdisciplinary pioneers<\/td><\/tr>
          Eigenvector<\/td>Influence weighted by neighbor importance<\/td>Core foundational status within an IP cluster<\/td>Distinguish true crown jewels from high-volume noise<\/td><\/tr>
          PageRank<\/td>Stationary probability of random citation walk<\/td>Global knowledge flow position<\/td>Rank-order entire portfolios for valuation; M&A screening<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

          Key Takeaways: Centrality Metrics<\/strong><\/p>\n\n\n\n

            \n
          • In-degree (forward citations) is the most intuitive value signal but must be normalized for patent age using citation velocity.<\/li>\n\n\n\n
          • Betweenness centrality is the most powerful tool for identifying convergence opportunities and under-explored white space at technology intersections.<\/li>\n\n\n\n
          • Eigenvector centrality separates true crown jewels from patents that are merely popular within a narrow niche.<\/li>\n\n\n\n
          • Using a single metric produces misleading rankings. A portfolio assessment should report all five measures and triangulate across them.<\/li>\n<\/ul>\n\n\n\n
            \n\n\n\n

            The Analyst’s Toolkit: Public Databases to Enterprise Platforms<\/strong><\/h2>\n\n\n\n

            The right tool depends on the question being asked, the budget available, and the technical depth of the analysis required.<\/p>\n\n\n\n

            Public Databases.<\/strong> Google Patents, the USPTO’s Patent Public Search, and the EPO’s Espacenet are the entry points for most patent work. They provide free access to the full text, claims, citations, and legal status of patents across major jurisdictions. Their limitations for network analysis are fundamental: the data is raw and unharmonized, citation records are inconsistently structured, and there are no integrated analytical capabilities. They are invaluable for document-level research and preliminary searches, but they are not viable platforms for portfolio-level citation network analysis.<\/p>\n\n\n\n

            Open-Source Visualization Tools.<\/strong> Gephi, VOSviewer, and CitNetExplorer are powerful network visualization engines that accept structured citation data as input and produce interactive network maps with full centrality calculations. VOSviewer uses the VOS (Visualization of Similarities) algorithm to place related nodes in spatial proximity, which produces visually intuitive cluster maps. CitNetExplorer is purpose-built for citation network analysis and includes temporal analysis features that are particularly useful for tracing technology trajectories. The critical limitation is that these tools do not source, clean, or structure data. The analyst is responsible for providing an analysis-ready input file, which requires all the harmonization work described above.<\/p>\n\n\n\n

            Commercial Intelligence Platforms.<\/strong> Clarivate’s Derwent Innovation and Innography, LexisNexis PatentSight, PatSnap, and Questel Orbit are the enterprise workhorses for corporate IP teams. Their core value proposition is the integration of three functions: global, multi-jurisdictional data coverage; proprietary harmonization of assignee data and corporate trees; and built-in analytical tools ranging from technology landscaping and trend analysis to citation network visualization. These platforms allow an analyst who is not a network scientist to generate high-quality citation maps and centrality calculations without building the data infrastructure from scratch. The cost of this convenience is substantial (enterprise licenses run to six or seven figures annually), but for an IP team running continuous competitive monitoring across a full therapeutic portfolio, the alternative cost in analyst time is higher.<\/p>\n\n\n\n

            Specialized Pharmaceutical Databases.<\/strong> DrugPatentWatch occupies a distinct and complementary position in this ecosystem. Where Clarivate and LexisNexis PatentSight focus on the IP layer, DrugPatentWatch focuses on the intersection of IP and pharmaceutical commercialization. It links patent records directly to FDA-approved drugs, Orange Book listings, patent expiry dates, patent term extensions, Paragraph IV challenge history, and biosimilar development pipelines. For a pharma strategist, this integration answers the question that enterprise IP platforms cannot: when exactly does this network’s most central patent expire, who has already challenged it, and which biosimilar applicants are in the queue? That question drives the decisions that matter most in pharmaceutical commercial strategy.<\/p>\n\n\n\n

            Table 2: Patent Analytics Resources Compared<\/strong><\/p>\n\n\n\n

            Resource Type<\/th>Primary Examples<\/th>Core Value<\/th>Data Harmonization<\/th>Analytical Depth<\/th>Best For<\/th><\/tr><\/thead>
            Public Databases<\/td>Google Patents, Espacenet, USPTO PPUBS<\/td>Full-text access, free<\/td>None (raw data)<\/td>Minimal<\/td>Individual document research, basic prior art<\/td><\/tr>
            Open-Source Visualization<\/td>Gephi, VOSviewer, CitNetExplorer<\/td>Powerful network visualization<\/td>User must supply clean data<\/td>High (if data is clean)<\/td>Academic research, custom one-off analyses<\/td><\/tr>
            Commercial IP Platforms<\/td>Clarivate Innography, LexisNexis PatentSight<\/td>End-to-end enterprise intelligence<\/td>High; core differentiator<\/td>Full suite<\/td>Corporate CI teams, continuous competitive monitoring<\/td><\/tr>
            Pharma-Specific Databases<\/td>DrugPatentWatch<\/td>Patent-to-drug commercial linkage<\/td>High within pharma context<\/td>Deep on expiry, litigation, biosimilar<\/td>BD, lifecycle management, Paragraph IV strategy<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
            \n\n\n\n

            Horizon Scanning: Detecting the Next Wave Before It Breaks<\/strong><\/h2>\n\n\n\n

            Horizon scanning is the systematic detection of early-stage technological signals that could reshape a therapeutic area or commercial market. Traditional approaches rely on expert opinion panels, key opinion leader (KOL) surveys, and regular reviews of the published literature. These methods are valuable but suffer from a common limitation: they are dependent on what experts are already paying attention to. The next major platform technology will not appear on a KOL survey until it is already being discussed at major conferences.<\/p>\n\n\n\n

            Patent citation network analysis provides an objective, data-driven complement to expert opinion. The methodology begins with a broad corpus of patents and, where relevant, scientific publications covering a therapeutic area of interest. This corpus is assembled using a combination of keyword queries, CPC classification codes, and a set of seed patents known to be central to the field. For a comprehensive analysis, the corpus should include both granted patents and published applications, capturing filings whose commercial trajectory is still undecided.<\/p>\n\n\n\n

            After constructing the citation network from this corpus, a community detection algorithm (typically Louvain or InfoMap) partitions it into distinct research clusters. The horizon scanning step involves characterizing the temporal properties of each cluster. Specifically, analysts calculate the median priority date of the patents within each cluster and the year-over-year growth rate of new filings added to each cluster. Clusters with a recent median priority date (within the past three to five years) and a high growth rate are the emerging research fronts.<\/p>\n\n\n\n

            To characterize the content of these emerging clusters, text mining methods extract the most distinctive terms. TF-ICF (Term Frequency-Inverse Cluster Frequency) is the preferred method: it identifies terms that appear frequently within a cluster but rarely across the full corpus, which produces a precise vocabulary that describes what is unique about the emerging field rather than what is generic to the broader therapeutic area.<\/p>\n\n\n\n

            A real-world application of this exact methodology to immunology citation data successfully identified ARID1A gene mutation as an emerging immuno-oncology target and CD300e as an emerging immune receptor of interest \u2014 both before these targets had attracted mainstream attention. The method works because researchers file patents early, often before peer-reviewed publication, and the citation relationships between early filings are structurally detectable in network analysis before the topic achieves broader recognition.<\/p>\n\n\n\n

            Practical Application: Three-Stage Horizon Scanning Process<\/strong><\/p>\n\n\n\n

            A production-grade horizon scanning process for a pharmaceutical R&D organization involves three stages, executed on a rolling basis:<\/p>\n\n\n\n

            Stage 1 is a broad sweep, conducted quarterly. The full corpus for a therapeutic area is refreshed, the network is rebuilt, and cluster growth rates are calculated. The output is a ranked list of the ten fastest-growing clusters by filing velocity, each characterized by its distinctive vocabulary and most central patents.<\/p>\n\n\n\n

            Stage 2 is a deep dive, triggered by any cluster appearing in the top three of two consecutive quarterly scans. An analyst prepares a detailed report on the cluster: who is filing (with harmonized assignee data), what the specific claims cover, what prior art the examiners are citing, and whether clinical trial registrations or scientific publications suggest the underlying biology is advancing.<\/p>\n\n\n\n

            Stage 3 is executive-level synthesis. Findings from Stage 2 are translated into R&D investment decisions, partnership targets, or licensing opportunities.<\/p>\n\n\n\n

            Key Takeaways: Horizon Scanning<\/strong><\/p>\n\n\n\n

              \n
            • Patent filings precede peer-reviewed publications by an average of 18-24 months in pharmaceutical research. Network-based horizon scanning captures signals that literature-based monitoring will miss for up to two years.<\/li>\n\n\n\n
            • Cluster growth rate and median priority date are the two most important parameters for identifying emerging research fronts.<\/li>\n\n\n\n
            • TF-ICF term extraction provides a reliable vocabulary for characterizing what is distinctive about a newly emerging cluster without requiring deep domain expertise in the specific sub-field.<\/li>\n<\/ul>\n\n\n\n
              \n\n\n\n

              Competitor Intelligence: Decoding Rival R&D from Patent Topology<\/strong><\/h2>\n\n\n\n

              A competitor’s patent portfolio is a structured record of their R&D decisions over time. Citation network analysis allows you to read that record with a precision that is not available from any other public source.<\/p>\n\n\n\n

              The first-order analysis is portfolio topology: what does the shape of a competitor’s patent network look like? A portfolio with one extremely high-in-degree patent surrounded by a dense cluster of lower-centrality patents is characteristic of a thicket strategy around a single platform asset. A portfolio with multiple high-centrality patents in separate clusters indicates a diversified innovation strategy across distinct technology areas. A portfolio dominated by patents with high out-degree but low in-degree suggests recent, active filing in a new area where the competitor is building knowledge but has not yet established foundational IP.<\/p>\n\n\n\n

              Filing Velocity Monitoring.<\/strong> The rate at which a competitor files in a specific CPC classification code is a leading indicator of R&D investment. A 40% increase in quarterly filings within a therapeutic area, detected six to twelve months before any public announcement, is a robust signal that the competitor is scaling up a program in that area. Automated alert systems on commercial platforms like DrugPatentWatch or Clarivate can be configured to trigger notification whenever a specified competitor files in a specified classification, bringing the signal to the analyst’s attention without requiring manual monitoring.<\/p>\n\n\n\n

              Geographic Expansion Signals.<\/strong> When a competitor begins filing national phase applications in a jurisdiction where they previously had no presence, they are signaling a new commercial priority. A Chinese pharmaceutical company’s first USPTO filing in a specific CPC class signals the beginning of a US market entry strategy. A US company’s first CNIPA filing in a specific therapeutic area signals serious attention to the Chinese market.<\/p>\n\n\n\n

              Detecting Stealth Programs.<\/strong> Not every R&D program gets announced in a press release or shown at a conference. Patent filing patterns often expose programs that companies are actively managing quietly. A coherent cluster of patents in a specific biological pathway, all assigned to the same company, all filed within the same 18-month window, constitutes a detectable signal of a systematic program even if the company has said nothing publicly about it. The signal is confirmed when the cluster’s most central patents have high examiner-added citation rates, indicating that the examiner’s independent search is surfacing the same prior art from multiple directions.<\/p>\n\n\n\n

              Citation Cross-Flows as Competitive Intelligence.<\/strong> The citation links that cross company boundaries provide a direct view of inter-company technological relationships. Firm A’s patents frequently citing Firm B’s core IP can mean one of three things: Firm A is building on Firm B’s technology and may owe royalties; Firm A is designing around Firm B’s claims and is disclosing the prior art to document its awareness; or Firm A’s examiner has cited Firm B’s patents in search reports. Determining which scenario applies requires reading the actual claims and file histories, but identifying which cross-company citation relationships exist at all is a powerful first filter for competitive intelligence work.<\/p>\n\n\n\n

              \n\n\n\n

              Evergreening and Patent Thickets: A Network-Level View<\/strong><\/h2>\n\n\n\n

              Evergreening and patent thicket construction are the two dominant secondary patent strategies in pharmaceutical IP management. Both are visible in citation network structure, and both require different analytical responses.<\/p>\n\n\n\n

              Evergreening<\/strong> involves filing secondary patents on modifications or extensions of an existing drug: new crystalline polymorphs, new salt forms, new dosing regimens, new methods of use for different indications, new formulations with improved bioavailability or stability, and new delivery devices. Each of these secondary patents, if granted, extends the period during which the innovator can seek Orange Book listing and assert infringement against ANDA filers. In a citation network, an evergreening strategy appears as a chain or fan of secondary patents clustered around a high-in-degree foundational API patent. The secondary patents will typically cite the original compound patent, creating an identifiable citation pattern.<\/p>\n\n\n\n

              The strategic implication for a generic developer is clear: the relevant question is not ‘when does the API patent expire?’ but ‘when does the last Orange Book-listed patent expire, and what is the validity risk of each listed secondary patent?’ DrugPatentWatch provides the litigation history and Paragraph IV challenge data needed to answer the second part of that question alongside the citation network structure that maps the first part.<\/p>\n\n\n\n

              Patent Thickets<\/strong> in biologics are structurally different. Rather than a linear extension of a foundational patent, a biologic thicket is a dense, multi-directional cluster of patents covering the active ingredient, its manufacturing process, its formulation, its delivery device, its dosing regimen, and its specific physiochemical properties (glycosylation patterns, aggregation characteristics, immunogenicity profiles). The thicket’s purpose is not primarily to extend exclusivity through sequential filings but to surround the product with enough valid patents that a biosimilar developer faces an overwhelming litigation burden on entry.<\/p>\n\n\n\n

              A 2023 study published in PMC reviewed all patents involved in US biologic litigation and found that while only 4% of litigated patents were primary patents on the active biological ingredient, 8% were ancillary product patents covering critical physiochemical properties, and the remaining 87% were non-ancillary secondary patents covering formulations, methods of use, and manufacturing. The ancillary patents were filed a median of 18.3 years after the first primary patent and extended expected protection by a median of 10.4 years. This data quantifies precisely what network analysis reveals visually: the thicket is the real barrier, and its most economically significant components are the ancillary patents, not the foundational ones.<\/p>\n\n\n\n

              For a biosimilar developer, overlaying the biologic patent network with litigation data from Lex Machina or Clarivate Darts-ip identifies which of the dozens of thicket patents the innovator has chosen to defend in court. These are the patents the innovator believes are valid, infringed, and worth litigating. This intersection of network analysis and litigation history is the most efficient tool available for prioritizing invalidity challenges and design-around efforts.<\/p>\n\n\n\n

              \n\n\n\n

              Biologic vs. Small Molecule IP Strategy: A Structural Comparison<\/strong><\/h2>\n\n\n\n

              The IP architecture for a major biologic and a major small molecule drug differs in ways that require distinct analytical frameworks. The following comparison is designed as a practical reference for IP teams working across both asset classes.<\/p>\n\n\n\n

              Table 3: IP Strategy Comparison \u2014 Biologics vs. Small Molecules<\/strong><\/p>\n\n\n\n

              Strategic Dimension<\/th>Small Molecules<\/th>Biologics \/ Large Molecules<\/th><\/tr><\/thead>
              Core patent focus<\/td>API as a distinct chemical entity (composition of matter claim)<\/td>Active ingredient plus manufacturing process; the process is integral to the product<\/td><\/tr>
              Secondary patent types<\/td>Polymorphs, salts, esters, pro-drugs, formulations, new methods of use, pediatric exclusivity<\/td>Formulations, delivery devices, dosing regimens, ancillary product properties, glycosylation patterns, manufacturing process parameters<\/td><\/tr>
              Primary competitive threat<\/td>ANDA filers under Hatch-Waxman Act; Paragraph IV certification<\/td>Biosimilar applicants under BPCIA; 351(k) pathway interchangeability designation<\/td><\/tr>
              Key exclusivity mechanism<\/td>Orange Book patent listing; Hatch-Waxman patent term extension; NCE exclusivity<\/td>BPCIA 12-year reference product exclusivity; 4-year data exclusivity; patent dance procedures under 42 U.S.C. 262(l)<\/td><\/tr>
              Dominant defensive strategy<\/td>Evergreening: sequential secondary patent filings to extend effective exclusivity<\/td>Thicket construction: dense, overlapping patent clusters to create litigation cost barriers<\/td><\/tr>
              Exclusivity duration driver<\/td>API patent expiry plus PTE extensions; polymorph\/formulation lifecycle management<\/td>Reference product exclusivity plus thicket-driven biosimilar delay; ancillary patents filed 15-20 years post-original filing<\/td><\/tr>
              Citation network topology<\/td>Hub-and-spoke: central API patent with radiating secondary filings<\/td>Dense multi-cluster: dozens of interconnected nodes with no single dominant hub<\/td><\/tr>
              Key analytical tool pairing<\/td>Network analysis + Orange Book data + Paragraph IV challenge history<\/td>Network analysis + Purple Book data + BPCIA patent dance filings + Lex Machina litigation data<\/td><\/tr>
              IRA price negotiation exposure<\/td>Small molecules selected for negotiation after 9 years post-approval<\/td>Biologics selected after 13 years post-approval; longer runway amplifies thicket investment rationale<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
              \n\n\n\n

              White Space Analysis: Finding Uncontested Innovation Ground<\/strong><\/h2>\n\n\n\n

              White space analysis maps the current IP landscape in a technology area to identify zones with low or absent patent coverage, representing potential opportunities for first-mover IP protection, differentiated product positioning, or unmet clinical needs that current R&D has not yet addressed.<\/p>\n\n\n\n

              The analysis begins with a clearly defined strategic question, not a broad topic. ‘What are the white spaces in oncology?’ will produce a map too diffuse to act on. ‘What biological pathways for overcoming acquired resistance to KRASG12C inhibitors have not been claimed in active patents?’ produces a map that generates specific, actionable R&D directions. The specificity of the input query determines the actionability of the output.<\/p>\n\n\n\n

              With a focused question defined, the corpus is assembled using a combination of CPC codes, keyword queries, and seed patents known to be central to the field. The assembled patents are structured as a citation network, clustered, and then visualized as a heat map or topographic density map. Dense, high-patent-count areas are the established competitive zones, typically dominated by a small number of large assignees. Sparse areas with few or no patents but active underlying science (visible in non-patent literature citations) are the candidate white spaces.<\/p>\n\n\n\n

              The critical analytical step after identifying candidate white spaces is the feasibility validation: why is this area empty? Four hypotheses should be evaluated for each candidate:<\/p>\n\n\n\n

              The area may be a ‘desert’ \u2014 technically infeasible and empty for a good scientific reason. Prior clinical failures, fundamental biological barriers, or ADMET liabilities may have deterred investment. This is confirmed by checking ClinicalTrials.gov for terminated or withdrawn trials in the area and reviewing the scientific literature for published failure analyses.<\/p>\n\n\n\n

              The area may be ‘pre-commercial’ \u2014 technically plausible but too early for commercial interest. Academic patent filings or preprint publications may be present without commercial assignees having followed. This represents a classic first-mover opportunity.<\/p>\n\n\n\n

              The area may be ‘claimed but invisible’ \u2014 patented by a subsidiary or under a name that the unharmonized query missed. This is exactly why data harmonization precedes white space analysis in any rigorous workflow.<\/p>\n\n\n\n

              The area may be a ‘strategic gap’ \u2014 intentionally left open by all players because the commercial market is too small, the regulatory path is too uncertain, or the patient population is too heterogeneous. This is confirmed by market analysis and KOL consultation.<\/p>\n\n\n\n

              Convergence White Space: The Most Valuable Form of Gap Analysis<\/strong><\/p>\n\n\n\n

              The most powerful form of white space analysis does not look for emptiness. It looks for unclaimed intersections between existing technology clusters. A network map with a dense cluster of ‘ADC (Antibody-Drug Conjugate) patents’ and a separate dense cluster of ‘CRISPR delivery mechanism patents’ may have a structural hole between them, a zone of low betweenness centrality and sparse cross-cluster edges, representing the unclaimed space of CRISPR-enhanced ADC targeting or CRISPR-enabled payload selection. These intersection white spaces are where platform technologies that combine two existing fields can generate high-centrality, high-value IP with strong defensibility.<\/p>\n\n\n\n

              Key Takeaways: White Space Analysis<\/strong><\/p>\n\n\n\n

                \n
              • Empty zones on a patent map require validation before they become R&D investment theses. The four hypotheses \u2014 desert, pre-commercial, invisible, and strategic gap \u2014 should be explicitly tested.<\/li>\n\n\n\n
              • The most valuable white spaces are typically at the intersection of two established technology clusters rather than in completely unoccupied territory.<\/li>\n\n\n\n
              • White space analysis conducted on unharmonized data frequently produces false positives: apparent gaps that are actually occupied by entities with non-obvious assignee names.<\/li>\n<\/ul>\n\n\n\n
                \n\n\n\n

                M&A Targeting and Due Diligence: IP Valuation in the Deal Room<\/strong><\/h2>\n\n\n\n

                Pharmaceutical M&A is, in most cases, an IP transaction. The acquirer is purchasing the rights to a drug’s regulatory exclusivity, its clinical data package, and its patent portfolio. Of these three assets, the patent portfolio has the longest duration and the most complex risk profile. A drug’s Phase III data is what it is; a patent’s value depends on its remaining life, its validity risk, its freedom-to-operate exposure, and its competitive positioning in the citation network.<\/p>\n\n\n\n

                Target Identification Through Network Analysis.<\/strong> A standard M&A screen might identify acquisition targets by market capitalization, therapeutic area focus, or pipeline stage. Network analysis adds a fourth dimension: IP positioning. An analysis of citation centrality across a therapeutic area may surface a company with a small market cap but a handful of patents that have unusually high in-degree and betweenness centrality. These patents may be foundational to an entire sub-field. Their small-cap owner may be a target that traditional financial screening would miss.<\/p>\n\n\n\n

                A practical screening workflow using network analysis runs as follows. First, identify the top 20 patents by in-degree centrality in the target therapeutic area. Second, identify the assignees of those patents that are not currently part of large incumbent portfolios. Third, run eigenvector centrality on those assignees’ full portfolios to confirm that the high-value patents are not outliers but reflect a consistently strong IP position. Fourth, overlay this with pipeline data to confirm clinical viability. The result is a ranked list of potential acquisition targets filtered by patent network quality rather than market cap.<\/p>\n\n\n\n

                Due Diligence Applications.<\/strong> Once a target is identified, patent network analysis contributes to four distinct components of the due diligence process.<\/p>\n\n\n\n

                Portfolio strength assessment uses network centrality metrics to rank the target’s patents by influence, identifies the ‘crown jewels’ that drive the acquisition thesis, and quantifies the thicket density around key commercial assets. This goes beyond a simple patent inventory to produce a structured view of where the value actually lives.<\/p>\n\n\n\n

                Freedom-to-operate risk mapping uses the network to identify the target’s most central patents’ backward citation dependencies on third-party IP. If the target’s key patents cite competitor or platform licensor patents as primary prior art, this can indicate royalty obligations or infringement exposure that materially affects valuation.<\/p>\n\n\n\n

                Validity risk assessment uses citation network data alongside litigation intelligence from Lex Machina, Darts-ip, and publicly available PTAB records to estimate the probability that each key patent survives challenge. High in-degree patents are more likely to attract invalidity challenges, but they are also more likely to have been examined thoroughly. Cross-referencing citation quality with prosecution history and the claims’ scope relative to the prior art produces a patent-by-patent validity risk score.<\/p>\n\n\n\n

                Financial valuation integration feeds network-derived quality scores into risk-adjusted NPV (rNPV) models. A patent with high eigenvector centrality, long remaining life, low validity risk, and no FTO encumbrances contributes materially more to an acquisition’s rNPV than a patent with the same face expiry date but poor network positioning and a pending IPR petition at PTAB.<\/p>\n\n\n\n

                The Innovation Culture Read.<\/strong> A network analysis of the target’s portfolio can reveal behavioral patterns that speak to R&D culture and post-merger integration prospects. A portfolio with high self-citation rates indicates either a coherent internal platform or an insular R&D organization. High NPL citation rates (particularly to recent academic literature) indicate strong connections to cutting-edge basic science, often correlated with active university licensing relationships or strong CRO partnerships. A pattern of high betweenness centrality patents suggests interdisciplinary R&D practices that may be difficult to replicate after the acquisition team disperses. These behavioral indicators are not captured in a standard patent inventory and require network analysis to surface.<\/p>\n\n\n\n

                Investment Strategy Note for Portfolio Managers and Analysts<\/strong><\/p>\n\n\n\n

                When evaluating a pharma or biotech company for investment, citation network quality metrics serve as independent validation of management’s R&D productivity claims. A company reporting strong pipeline progress but with a patent portfolio showing declining network centrality over the past three years is showing a disconnect that warrants investigation. Conversely, a company in a perceived ‘late lifecycle’ phase but whose patent network shows increasing betweenness centrality across new technology clusters may be executing a silent platform pivot that financial models are not yet pricing in. Citation velocity trends over 36-month rolling windows provide a quantitative basis for these judgments.<\/p>\n\n\n\n

                \n\n\n\n

                Technological Convergence: Mapping Where Disciplines Collide<\/strong><\/h2>\n\n\n\n

                The most commercially significant pharmaceutical innovations of the next decade will come from the intersection of biology, chemistry, data science, and medical device engineering. CAR-T therapies combine gene editing, immunology, and cell manufacturing. AI-discovered small molecules combine computational chemistry, machine learning, and traditional medicinal chemistry. Digital therapeutics combine behavioral science, software engineering, and clinical evidence generation. Patent citation networks are uniquely suited to detecting these convergences early, because the citations that cross CPC classification boundaries are exactly the citations that map inter-disciplinary flows.<\/p>\n\n\n\n

                The analytical approach is straightforward once the network is built with multi-classification tagging. Each patent carries CPC codes across the hierarchical classification system. A patent with primary classification in A61K (pharmaceutical preparations) and secondary classifications in G16H (healthcare informatics) and G06N (machine learning) is, by its classification structure, a document at the intersection of pharma and AI. In a citation network, patents like this that also have high betweenness centrality \u2014 sitting on the paths between a cluster of pharma patents and a cluster of AI patents \u2014 are the empirical evidence that these fields are converging at a commercially significant level.<\/p>\n\n\n\n

                Tracking the density of cross-cluster citation links over time provides a convergence timeline. If, in 2019, there were 50 cross-citations between the oncology patent cluster and the AI\/ML patent cluster in a given dataset, and by 2024 there are 850, that is not a trend; that is a structural transformation underway. The companies with the most patents in the high-betweenness intersection zone at the right time will own the platform IP that all subsequent practitioners must cite and license.<\/p>\n\n\n\n

                A landmark landscape analysis of personalized medicine patents mapped by ResearchGate found exactly this pattern: traditional therapeutic clusters in oncology, neurodegenerative diseases, and infectious diseases all showing accelerating cross-cluster linkages with IT-driven diagnostics and data science patents. The implication for pharmaceutical companies is specific: the first movers to file high-centrality patents at these intersections will define the IP boundaries of computational medicine. The companies that wait for clinical proof of concept before filing will be writing dependent claims, not foundational ones.<\/p>\n\n\n\n

                \n\n\n\n

                Oncology as a Case Study: The Densest Patent Battlefield in Medicine<\/strong><\/h2>\n\n\n\n

                Oncology is the most patent-intensive, most litigated, and most AI-affected therapeutic area in global pharmaceutical R&D. It provides an ideal environment to demonstrate all the frameworks described in this guide simultaneously.<\/p>\n\n\n\n

                Between 2015 and 2021, patenting in oncology-related technologies increased by over 70%, according to the EPO’s 2023 study ‘Patents and Innovation Against Cancer.’ The United States accounts for approximately 45% of international oncology patent families. The field spans an unusually wide technological range: traditional small molecule kinase inhibitors, monoclonal antibodies and bispecific antibodies, ADCs, CAR-T and TCR-T cell therapies, cancer vaccines (both prophylactic and therapeutic), RNA-based approaches (mRNA, siRNA, ASOs), AI-driven biomarker discovery, and companion diagnostic platforms. Each of these sub-fields constitutes a distinct cluster in a citation network analysis, each with its own center-of-gravity patents, its own leading assignees, and its own developmental timeline.<\/p>\n\n\n\n

                Immuno-Oncology as a Network Study.<\/strong> The PD-1\/PD-L1 checkpoint inhibitor space illustrates how network analysis maps a rapidly evolved IP landscape. The foundational science patents in this space, originating from Tasuku Honjo’s lab at Kyoto University and Gordon Freeman’s work at Dana-Farber Cancer Institute, now carry extraordinarily high in-degree centrality: they have been cited by hundreds of subsequent patents on combination therapies, biomarkers, dosing optimization, and patient selection. Bristol-Myers Squibb (nivolumab), Merck & Co. (pembrolizumab), and Roche (atezolizumab) have each built distinct patent clusters around these foundational nodes. Mapping the citation relationships between these three clusters \u2014 the cross-company citation flows \u2014 reveals which companies are building on each other’s work versus pursuing divergent technological paths, which in turn informs freedom-to-operate analysis and licensing negotiation positioning.<\/p>\n\n\n\n

                Drug Discovery Constraints as Network Phenomena.<\/strong> A 2020 study published in PMC (‘Can Literature Analysis Identify Innovation Drivers in Drug Discovery?’) identified two structural features of drug discovery R&D that are directly visible in patent networks: ‘preferential attachment,’ the tendency to cluster research around already highly-studied targets, and ‘local network effects,’ the tendency to explore only proteins that physically interact with well-established targets. In oncology, these phenomena mean that the PD-1\/PD-L1, EGFR, HER2, VEGF, and KRAS spaces are among the most crowded areas in the entire citation network, while large portions of the human kinome, proteome, and epigenome remain underexplored. A white space analysis overlaid on the oncology citation network makes these structural biases visible and quantifiable, providing an objective basis for a contrarian R&D strategy.<\/p>\n\n\n\n

                CAR-T and the Emerging IP Battleground.<\/strong> The CAR-T therapy patent landscape illustrates convergence analysis in real time. Citation networks built from CAR-T patents show three converging clusters: an immunology and T-cell biology cluster rooted in foundational academic work; a gene editing cluster anchored by CRISPR and TCR engineering patents; and a cell manufacturing and process scale-up cluster. The patents with highest betweenness centrality in this emerging space will be the most commercially significant over the next decade, because they bridge the three disciplines that must all work together for CAR-T to scale to widespread clinical use. Novartis, Kite\/Gilead, Bristol-Myers Squibb, and a cohort of university licensees (particularly Penn Medicine and City of Hope) dominate the current network, but the convergence cluster at the intersection of CRISPR and manufacturing process patents is significantly less concentrated, representing a viable entry point for companies with the right platform capabilities.<\/p>\n\n\n\n

                \n\n\n\n

                Biologics and the BPCIA Battlefield: Ancillary Patents and Thicket Architecture<\/strong><\/h2>\n\n\n\n

                Biologics account for approximately 2% of prescriptions in the US but represent 50% of total prescription drug spending. This extreme revenue concentration makes protecting biologic assets with every available IP tool an absolute business imperative. The strategic response from innovator companies has been the development of sophisticated thicket architectures that have delayed meaningful biosimilar competition for multiple major products beyond what any single patent expiry date would predict.<\/p>\n\n\n\n

                The BPCIA Patent Dance.<\/strong> The Biologics Price Competition and Innovation Act of 2010 established the 351(k) biosimilar approval pathway and created a structured information exchange process between innovator and biosimilar applicants, known informally as the ‘patent dance.’ Under 42 U.S.C. 262(l), a biosimilar applicant must provide its application and manufacturing information to the reference product sponsor, who then identifies all patents that could be asserted and proposes a list for litigation. The biosimilar applicant responds with its non-infringement or invalidity positions, and the parties negotiate a list of patents for immediate litigation. This process, while designed to encourage early patent dispute resolution, also hands innovator companies information about the biosimilar applicant’s manufacturing process before any public disclosure, which they can use to inform additional patent filings \u2014 a strategic advantage that the patent dance procedurally enables.<\/p>\n\n\n\n

                Thicket Architecture and Ancillary Patent Strategy.<\/strong> The landmark PMC study cited above provides the quantitative basis for understanding how thickets work in biologics litigation. The median ancillary product patent (covering critical physiochemical properties like glycosylation patterns, aggregation behavior, and immunogenicity profiles) was filed 18.3 years after the first primary patent on the active biological ingredient, extending expected protection by a median of 10.4 years. These are not incremental additions; they are the structural backbone of the long-term exclusivity strategy.<\/p>\n\n\n\n

                For a biosimilar developer, the analytical implication is that a patent network map of the reference product’s IP must include these ancillary patents as primary targets for invalidity strategy, even though they are secondary patents by definition. The question to answer for each ancillary patent is: does this patent cover a feature of the reference product that the biosimilar product structurally cannot avoid replicating? If yes, it is a critical barrier that requires either a successful invalidity challenge (likely at PTAB via IPR petition) or a license. If no, it is a barrier that can be designed around.<\/p>\n\n\n\n

                Reference Product Case Example: Humira (adalimumab).<\/strong> AbbVie’s adalimumab patent estate is the most-analyzed thicket in biologic IP history. Over 100 patents cover various aspects of the product, including the antibody composition, formulation, dosing regimen, delivery device, and manufacturing process. The last of these patents does not expire until the mid-2030s in some jurisdictions. Biosimilar competition has been substantially more robust in Europe, where AbbVie chose not to build out the same thicket depth it deployed in the US, and where the reference product exclusivity provisions of the BPCIA do not apply. The citation network for Humira-related patents, when mapped by jurisdiction, shows dramatically different thicket density between US and European patent families, illustrating how market-specific IP strategy shapes the competitive entry timeline.<\/p>\n\n\n\n

                IP Valuation Note for Humira Biosimilar Portfolio Managers.<\/strong> As of 2026, multiple adalimumab biosimilars have cleared US regulatory approval (Amjevita, Cyltezo, Hadlima, Hyrimoz, Hulio, Simlandi, Yuflyma). Their market penetration trajectories have differed based on high-concentration\/low-volume formulation distinctions, interchangeability designations, and formulary contracting dynamics, not just IP clearance. This illustrates a broader principle: citation network analysis can map the IP barriers, but commercial success post-biosimilar entry depends on factors that patent data alone cannot predict.<\/p>\n\n\n\n

                \n\n\n\n

                Limitations and Biases: Where Citation Analysis Can Mislead You<\/strong><\/h2>\n\n\n\n

                Patent citation networks are powerful analytical tools, but they are not a neutral or complete representation of the innovation landscape. Five categories of limitation require explicit acknowledgment and methodological responses.<\/p>\n\n\n\n

                Truncation Bias.<\/strong> A patent filed in 2023 will have had less time to accumulate forward citations than a patent filed in 2013, regardless of their relative importance. Raw citation counts systematically undervalue recent patents relative to older ones. Citation velocity (citations per year) partially corrects this, but it does not fully eliminate the bias in rapidly evolving fields where the most influential patents generate citations in the first two or three years. For any analysis that compares patents across different filing cohorts, age-normalized metrics are mandatory.<\/p>\n\n\n\n

                Strategic Over-Citing.<\/strong> To ensure compliance with the duty of disclosure while minimizing the risk of identifying any single piece of prior art as critical, applicants may submit long lists of citations that they consider marginally material. This ‘shotgun disclosure’ practice inflates citation counts for documents that appear in many such lists without being genuinely foundational. The noise this introduces is systematic and difficult to remove algorithmically. Weighting examiner-added citations more heavily than applicant-added citations, as discussed above, is the most effective mitigation.<\/p>\n\n\n\n

                Strategic Under-Citing.<\/strong> Less common than over-citing but more analytically damaging: applicants may omit citations to the most directly relevant prior art, particularly prior art from foreign jurisdictions or art that would most severely threaten their claims. Patent applications filed in this pattern will overestimate the novelty of the invention and create false white space in the network. The examiner’s search report provides a check on this, but examiners cannot be assumed to catch everything. Cross-referencing the citation record with a manual prior art search in important sub-fields is good analytical hygiene.<\/p>\n\n\n\n

                Citation as Imperfect Value Proxy.<\/strong> Forward citations correlate positively with patent value, but the correlation is not strong enough to use as a sole valuation metric. High commercial value sometimes accrues to patents that cover a narrow but practically inescapable feature of a best-selling product, rather than a scientifically generative foundational invention. These ‘choke point’ patents may have modest citation counts because they cover a very specific commercial embodiment. Conversely, some highly-cited patents cover foundational scientific concepts that turn out to have limited commercial applicability.<\/p>\n\n\n\n

                Legal Uncertainty Risk.<\/strong> A patent’s legal status is always subject to change. The Supreme Court’s 2023 decision in Amgen Inc. v. Sanofi<\/em>, which invalidated Amgen’s antibody claims for lack of enablement, has added substantial uncertainty to the valuation of broad biologic platform patents. Any patent network analysis that values assets based on the assumption that broad functional claims will survive post-grant challenge needs to be revisited in the light of this decision. The PTAB IPR (Inter Partes Review) process continues to invalidate a significant fraction of challenged claims. Litigation intelligence data, overlaid on the citation network, is the only way to account for this uncertainty systematically.<\/p>\n\n\n\n

                Emerging Technology Patentability.<\/strong> AI-generated drug candidates, CAR-T constructs designed by machine learning, and predictive diagnostics based on proprietary models all face contested patentability standards. Current USPTO guidance requires that all claimed inventions have ‘significant human contribution’ to a natural person inventor. What constitutes ‘significant’ is being actively litigated and may change materially. Network analysis of AI-generated drug patents should incorporate a validity risk discount until the patentability standards for these assets are settled by the courts or by congressional action.<\/p>\n\n\n\n

                \n\n\n\n

                AI-Powered Patent Analytics: The Emerging Infrastructure<\/strong><\/h2>\n\n\n\n

                AI is simultaneously the most disruptive force in drug discovery and the most transformative technology entering the patent intelligence stack. Understanding both dimensions is necessary for anyone managing pharmaceutical IP in 2026.<\/p>\n\n\n\n

                AI in the Discovery Pipeline.<\/strong> Preclinical programs using machine learning for target identification, hit-to-lead optimization, and ADMET prediction are now routinely claiming 30-50% reductions in time-to-candidate selection relative to fully wet-lab approaches. Insilico Medicine’s clinical-stage program for idiopathic pulmonary fibrosis (INS018_055) and Recursion Pharmaceuticals’ collaboration pipeline provide empirical benchmarks. The patent implications are immediate: AI-assisted discovery compresses the timeline from scientific insight to patent filing, potentially reducing the window between emergence in citation network analysis and commercial candidacy announcement.<\/p>\n\n\n\n

                AI in Patent Analytics.<\/strong> The integration of large language models and machine learning into patent intelligence platforms is producing specific capabilities that change the economics of analysis:<\/p>\n\n\n\n

                Semantic search (using transformer-based embeddings rather than keyword matching) produces substantially more complete and precise results for prior art identification. A semantic query for ‘antibodies that block checkpoint inhibition through a steric mechanism’ will surface relevant prior art that keyword queries miss because they require exact terminology. This capability reduces the risk of the ‘omitted prior art’ problem described in the limitations section above.<\/p>\n\n\n\n

                Automated classification and clustering using neural networks can process new patent filings in near real-time and assign them to the correct technology cluster, updating the citation network map continuously rather than in periodic analytical cycles. This enables the kind of real-time competitive monitoring that was previously only feasible for a narrow set of manually tracked competitors.<\/p>\n\n\n\n

                Predictive value scoring, where ML models trained on historical patent outcomes (citation trajectories, litigation outcomes, commercial deal values) produce forward-looking value estimates for new filings, has shown R-squared values above 0.40 in early research. This is not sufficient precision for individual asset valuation, but it is useful as a screening tool to prioritize which patents warrant deeper analyst attention.<\/p>\n\n\n\n

                The AI Inventorship Problem and Its Valuation Consequences.<\/strong> Any IP strategy built on AI-assisted or AI-directed drug discovery must grapple with the inventorship question. Current US law requires that inventors be natural persons. The USPTO’s 2024 guidance clarifies that human contribution to an AI-assisted invention must be ‘significant’ to qualify for patent protection, but it does not define ‘significant’ with the precision that a patent examiner or a court will require when the claim is challenged. This creates a category of AI-related pharmaceutical patents that carry elevated validity risk until the legal standard is clarified through litigation or legislation. Portfolio managers and acquirers should apply a validity risk discount to any pharma patent in which AI played a role in claim conception, sized according to the breadth of the claims and the thinness of the documented human contribution.<\/p>\n\n\n\n

                \n\n\n\n

                Investment Strategy for Portfolio Managers and Analysts<\/strong><\/h2>\n\n\n\n

                This section synthesizes the frameworks above into specific, actionable guidance for institutional investors and equity analysts covering pharmaceutical and biotechnology companies.<\/p>\n\n\n\n

                Patent Network Quality as a Forward Indicator of R&D Productivity.<\/strong> Academic economics research, including Trajtenberg’s foundational 1990 study on CT scanner patents, has established that citation-weighted patent counts are materially better predictors of R&D value than raw patent counts. For equity analysts, this means that tracking the citation velocity trend of a company’s patent portfolio over rolling 36-month windows provides an independent signal of R&D productivity that precedes clinical and financial announcements by 18-24 months.<\/p>\n\n\n\n

                Flags for Long and Short Positions.<\/strong> A company with a declining citation velocity trend \u2014 its recent filings are accumulating citations at a slower rate than its earlier filings \u2014 is showing IP-level evidence of slowing innovation. If this trend is not reflected in consensus pipeline projections or valuation multiples, it represents a potential short catalyst. Conversely, a company with rapidly increasing citation velocity in an emerging cluster, particularly if the cluster shows convergence characteristics connecting two previously separate technology areas, may be building a platform IP position that its current market cap does not reflect.<\/p>\n\n\n\n

                Thicket Depth as a Defensive Moat Metric.<\/strong> For established large-cap pharma positions, thicket depth and the remaining life of Orange Book and Purple Book-listed secondary patents are the most direct quantitative measures of the durability of a product’s exclusivity. Investors in companies with major biologic or blockbuster small molecule franchises should track the secondary patent filing activity and validity risk profiles for these products, using tools like DrugPatentWatch alongside network centrality analysis, to monitor whether the moat is widening or narrowing relative to biosimilar and generic competitive activity.<\/p>\n\n\n\n

                M&A Screening Using Network Data.<\/strong> The patent network screen for acquisition targets described in the M&A section above is directly applicable to investment screening. Small-cap and mid-cap biotechs with network-central patents in high-growth technology clusters are the most likely M&A targets, because they are the companies that large-cap incumbents need to acquire to maintain their network positions. Identifying these companies before acquisition announcements, using citation network screening combined with pipeline analysis, is a structured approach to identifying pre-announcement acquisition premium candidates.<\/p>\n\n\n\n

                Regulatory and Legal Event Risk.<\/strong> IPR petitions filed at PTAB against a company’s most central patents represent a quantifiable event risk that is systematically underfollowed by generalist equity coverage. A successful IPR that invalidates a foundational patent can immediately alter the competitive dynamics for a product category. Combining PTAB petition monitoring with patent network centrality data identifies which petitions, if successful, would have the highest impact on a given company’s IP moat.<\/p>\n\n\n\n

                \n\n\n\n

                Key Takeaways by Segment<\/strong><\/h2>\n\n\n\n

                For IP Teams and Patent Counsel<\/strong><\/p>\n\n\n\n

                  \n
                • Build citation analysis into the prosecution strategy. Examiner-added citations are objective competitive intelligence; review them systematically, not just as procedural documentation.<\/li>\n\n\n\n
                • Centrality metrics provide an objective basis for patent portfolio prioritization decisions. Not every asset warrants the same maintenance investment.<\/li>\n\n\n\n
                • The distinction between applicant-added and examiner-added citations should be captured and tracked as a data field in your internal patent management system.<\/li>\n<\/ul>\n\n\n\n

                  For R&D Leadership<\/strong><\/p>\n\n\n\n

                    \n
                  • Horizon scanning via citation network analysis provides a 12-24 month signal advantage over literature-based monitoring for emerging research fronts.<\/li>\n\n\n\n
                  • White space analysis at technology cluster intersections (high betweenness zones) is the most reliable method for identifying genuinely novel platform IP opportunities.<\/li>\n\n\n\n
                  • The topology of competitors’ portfolios reveals program priorities, technology dependencies, and strategic gaps that do not appear in press releases or conference presentations.<\/li>\n<\/ul>\n\n\n\n

                    For Business Development and Licensing<\/strong><\/p>\n\n\n\n

                      \n
                    • Citation velocity and eigenvector centrality are the two most actionable metrics for prioritizing inbound licensing opportunities and outbound deal targeting.<\/li>\n\n\n\n
                    • Cross-company citation flow analysis identifies both potential licensing partners (companies whose technologies you cite heavily) and potential infringement risks (companies that cite your core IP from directions you have not monitored).<\/li>\n<\/ul>\n\n\n\n

                      For Institutional Investors and Portfolio Managers<\/strong><\/p>\n\n\n\n

                        \n
                      • Citation-weighted patent counts are empirically validated predictors of R&D value, outperforming raw patent counts with correlation coefficients reaching 0.75 in foundational academic work.<\/li>\n\n\n\n
                      • Declining citation velocity in a company’s recent filing cohort, compared against historical trend, is a leading indicator of slowing R&D productivity that precedes clinical disappointments.<\/li>\n\n\n\n
                      • Thicket depth and secondary patent validity risk are measurable moat metrics for established biologic and blockbuster franchises.<\/li>\n<\/ul>\n\n\n\n
                        \n\n\n\n

                        FAQ: Practitioner Questions Answered<\/strong><\/h2>\n\n\n\n

                        Q: Our biotech has a limited analytics budget. What is the minimum viable approach to patent network analysis?<\/strong><\/p>\n\n\n\n

                        Define a single, focused competitive question (e.g., ‘Who owns the foundational patents in STING agonist design, and when do they expire?’). Use Google Patents to collect the relevant documents and their citations manually. Clean the assignee data in a spreadsheet by hand, which is feasible for a focused dataset of 200-500 patents. Import the clean citation matrix into the free version of VOSviewer. The resulting network map, combined with DrugPatentWatch data on expiry dates and litigation history, will produce actionable intelligence within a few days of analyst time. This is not a substitute for continuous enterprise monitoring, but it answers a specific strategic question at near-zero cost.<\/p>\n\n\n\n

                        Q: How do we distinguish a genuine white space from a technology desert when both look like absence on a patent map?<\/strong><\/p>\n\n\n\n

                        Test four hypotheses explicitly: scientific feasibility (is there positive NPL or clinical data suggesting the approach could work?), prior clinical failure (check ClinicalTrials.gov for terminated trials in the area), hidden occupancy (run the same query with unharmonized data and compare \u2014 if new entities appear in the unharmonized run, investigate their corporate trees), and commercial rationale absence (KOL consultation and market sizing to determine whether patient population or reimbursement dynamics deter investment). A genuine white space opportunity clears all four tests. A desert fails the first one.<\/p>\n\n\n\n

                        Q: How should we value AI-discovered drug patents given the legal uncertainty around AI inventorship?<\/strong><\/p>\n\n\n\n

                        Apply a validity risk discount to any patent where AI played a role in claim conception. The discount should be sized based on the breadth of the claims and the quality of the documented human contribution. Broad functional claims with thin human contribution in AI-assisted programs face the highest risk under the post-Amgen v. Sanofi<\/em> enablement standard and the USPTO’s inventorship guidance. Narrow, species-level claims with detailed experimental validation performed by identified human inventors face lower risk. Until US case law or legislation clarifies what constitutes ‘significant’ human contribution, any valuation or deal structure for an AI-assisted biotech asset should include this risk explicitly in the rNPV model.<\/p>\n\n\n\n

                        Q: How do we use NPL (non-patent literature) citations alongside patent citations in network analysis?<\/strong><\/p>\n\n\n\n

                        Build a bipartite network that includes both patent nodes and NPL document nodes. Citation links from patents to journal articles, and from journal articles to subsequent patents (as tracked through forward citation databases like Scopus or Web of Science), create a richer topological map that shows the linkage between applied IP and basic science. Companies whose patents cite recent high-impact publications (high NPL citation rates to papers published within the past five years) are demonstrating deep engagement with the current state of basic research. Companies with low NPL citation rates may be operating at greater distance from the scientific frontier. For M&A due diligence, the NPL citation profile of a target’s portfolio is a useful indicator of its connection to cutting-edge academic science and the depth of its scientific advisory relationships.<\/p>\n\n\n\n

                        Q: Can citation network analysis predict whether a patent will face an IPR challenge at PTAB?<\/strong><\/p>\n\n\n\n

                        With increasing reliability, yes. Research published in PMC in 2025 (‘Predicting patent challenges for small-molecule drugs: A cross-sectional study’) found that market size is a stronger predictor of patent challenge than citation count, consistent with the economic logic that patent challenges are strategically motivated rather than purely merit-based. However, network-derived validity signals \u2014 particularly the number and quality of examiner-added citations to close prior art, the breadth of claims relative to enablement support, and the presence of active PTAB petitions against related family members \u2014 are independent predictors of challenge probability. A commercial litigation intelligence platform (Lex Machina, Darts-ip) integrated with citation network centrality data provides the most complete picture available for assessing IPR exposure.<\/p>\n","protected":false},"excerpt":{"rendered":"

                        Why Standard Patent Intelligence Is Failing You The patent cliff is no longer the right mental model. For two decades, […]<\/p>\n","protected":false},"author":1,"featured_media":35426,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_lmt_disableupdate":"","_lmt_disable":"","site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"default","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","ast-disable-related-posts":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"footnotes":""},"categories":[10],"tags":[],"class_list":["post-34528","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-insights"],"modified_by":"DrugPatentWatch","_links":{"self":[{"href":"https:\/\/www.drugpatentwatch.com\/blog\/wp-json\/wp\/v2\/posts\/34528","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.drugpatentwatch.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.drugpatentwatch.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.drugpatentwatch.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.drugpatentwatch.com\/blog\/wp-json\/wp\/v2\/comments?post=34528"}],"version-history":[{"count":4,"href":"https:\/\/www.drugpatentwatch.com\/blog\/wp-json\/wp\/v2\/posts\/34528\/revisions"}],"predecessor-version":[{"id":38200,"href":"https:\/\/www.drugpatentwatch.com\/blog\/wp-json\/wp\/v2\/posts\/34528\/revisions\/38200"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.drugpatentwatch.com\/blog\/wp-json\/wp\/v2\/media\/35426"}],"wp:attachment":[{"href":"https:\/\/www.drugpatentwatch.com\/blog\/wp-json\/wp\/v2\/media?parent=34528"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.drugpatentwatch.com\/blog\/wp-json\/wp\/v2\/categories?post=34528"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.drugpatentwatch.com\/blog\/wp-json\/wp\/v2\/tags?post=34528"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}