Listen to this article
Artificial intelligence is already making drug development faster. Could it be used to find and exploit drug patent loopholes?
Already, AI is being used to extract, analyze, and manage data – data which may or may not be directly related to synthesis of drug molecules. With AI, drug developers can tease apart insights from unstructured data like scientific papers, analyze huge quantities of never-before-connected data to find hidden drug-disease correlations, and identify compounds which are likeliest to make it to market.
Pharmaceutical startups are using AI to bring their R&D in-house, because of the cost savings AI confers. And now, artificial intelligence is being used to innovate molecules through multiple synthetic chemical pathways. The potential is there both to help developers skirt existing drug patents through novel synthetic pathways, and to help drug developers protect their own patents by tightening up the very loopholes that novel synthetic pathways represent.
Generating Synthetic Sequences to Create Target Molecules
Chematica is an AI computer code/database combination that was originally developed in Poland. In 2017, it was bought by the Merck Group. The software is designed to combine long synthesis pathways into shorter, more economical pathways, and to find alternate synthetic pathways for the creation of existing molecules.
Achieving this requires an enormous knowledge base, along with pattern recognition algorithms, and perfect “memory” of reactions that have been tried. This combination of needs is exactly the sort of problem AI is designed for.
Without AI, the process of finding new synthetic pathways to the same molecule is painstaking and time-consuming. With Chematica, however, one team was able to generate synthetic routes for the creation of eight pre-defined molecules within 15 to 20 minutes each. The team then tested the results by following the synthetic routes to ensure the program created the desired target, and it did so in all eight cases.
Chemists have reproduced the synthetic pathways generated by AI programs to demonstrate that they actually work.
Using AI to Challenge Drug Patents
It’s easy to see how this form of AI could be used to skirt drug patents. A developer could simply use Chematica to explore all possible synthetic routes to the target molecule, and then create it using a pathway that the original drug developer did not patent.
New drug molecules are typically protected by numerous individual drug patents, hence the term “patent fence.” These various patents often cover multiple processes by which the drugs are synthesized, and they have traditionally been effective at keeping competitors away.
Artificial intelligence tools like Chematica, however, could allow competitors to develop synthetic routes the original manufacturer never conceived of. Drug manufacturers will have to be even more careful regarding compound, composition, and method of use patents going forward than they already are.
Using AI to Strengthen Drug Patents
Of course, there’s no reason pharmaceutical developers themselves can’t use tools like Chematica to develop (and patent) multiple synthesis processes, making the patent fence that much more impenetrable. When AI software has access to huge databases of millions upon millions of reactions, it can pick out “best routes” for the synthesis of drug molecules that even the most experienced chemists would not have come up with, simply based on the more massive quantity of data to which it has access.
What could end up happening is an “arms race” between drug companies, who want to cover as many bases as possible with their patents, and competitors, who want to find synthetic pathways that are not covered by existing patents – something that the astonishing power of machine learning makes possible.
Ultimately, artificial intelligence is a tool, and its use in the development of new drugs is already resulting in innovations and cost-saving methods. Things get a bit stickier when AI is applied to analyzing chemical routes to patented drug molecules, finding loopholes in those patents, and slipping through them. One sure bet is that drug manufacturers will be using the same AI technology to try to locate those loopholes early on, and close them via drug patents before competitors can exploit them.