Drug developers should publish fewer early findings because of AI, says investor Bruce Booth
Drug developers should publish fewer early findings because of AI, says investor Bruce Booth
On October 30, Bruce Booth, a partner at investment firm Atlas Venture, predicted lower returns on overall investment in drug development over the next three to five years. In his scenario, AI will reduce the cost of designing molecules for established therapeutic approaches, leading developers to produce more similar drugs.
Developers will have to compete for laboratory capacity, production of experimental batches, and trial participants. In Booth’s model, each company can choose its best project, while the industry as a whole pays to test many projects in parallel. Some new drugs may offer patients greater benefits than existing treatments; comparing their effects in trials takes years.
According to Booth, publications on new molecules help competitors move faster. Competitors take a published molecular structure and use AI to search for similar variants with the desired therapeutic effect. Their aim is to secure their own patent on a variant that falls outside the scope of an existing patent. The faster this search becomes, the sooner the original team faces rivals developing similar drugs.
In the US, a patent covering an entire class of molecules requires disclosure of how to make and use molecules across the full claimed class. In the dispute between pharmaceutical companies Amgen and Sanofi, the US Supreme Court upheld the invalidation of Amgen’s broad patent protection in 2023. Amgen had described 26 antibodies, protein molecules that block a protein that hinders the removal of cholesterol from the blood. The company sought protection for many more variants with the same effect. The Court found the disclosure insufficient to support that scope: scientists would have had to conduct an extensive new laboratory search to identify suitable molecules, requiring too much experimentation.
In April, biotech entrepreneur Paul Kohlhaas disclosed the molecular structure of a drug candidate and proposed competing on the speed of laboratory testing. Booth advises pioneers to keep their findings as trade secrets for longer by filing patent applications later and presenting early projects at conferences less often. His recommendation for scientific papers is to
“Limit the number and depth of scientific publications.”The longer findings remain secret, the later other researchers can use them in their work.
Booth cites the pharmaceutical company Eli Lilly and Company as an example. Its pipeline page, with data as of October 29, displays a notice that it has stopped disclosing Phase 1 projects in place of an overall list of those projects. Lilly defines this stage as beginning with the first dose in a human study.
In a response to the article, discussion participant Elly Shin proposes applying AI to the work of laboratories and manufacturers of experimental drugs so that more ideas can be tested:
“Make the drug development infrastructure itself cheaper, faster, and better.”