The Pistoia Alliance’s annual London conference has become a reliable read on what direction the life sciences community is heading, and the focus over the two days in London was on data quality, AI readiness and what practical deployment actually looks like: three sides of the same conversation.
Solving the data problem
Several presenters were blunt about a familiar problem: layering AI and AI agents on top of fractured, siloed data doesn’t accelerate discovery. It just produces faster chaos.
FAIR data principles, and the interoperability question in particular, kept surfacing as the reason why. Without a common language between systems, every integration is a custom build, and every AI initiative starts from scratch.
Sessions led by Elsevier and AstraZeneca put that in sharp relief. For AstraZeneca, the problem was concrete: CMC data spread across six separate platforms, pulled manually, regularly producing conflicting results. A centralized hub that harmonizes data at source changed that picture and made meaningful AI deployment possible.
Elsevier’s perspective was complementary: harmonizing internal data is only part of the challenge. Connecting it reliably to the broader scientific record, with the ontologies, retrieval methods and human curation that trustworthy AI requires, is where the real complexity lives.
The impact of practical AI
Sapio’s Rob Brown ran a session showing what becomes possible when the data foundation is in place. ELNs have gone through a number of “generations” over the last 20 years. First replacing paper, then making data easier to retrieve and share. But both were essentially passive.
Rob spoke about Sapio ELN, incorporating Sapio Elain, as the third generation, an active co-scientist that understands the experiment, coordinates the tools and helps the scientist decide what to do with the results.
Rob’s demo ran a live workflow entirely from within the ELN: synthetic accessibility scoring via Reaxys, retrosynthetic analysis on the selected molecule, and automatic creation of a full med-chem experiment, reactants and reagents included. One environment, no manual transfers.
Rob closed by connecting Sapio Elain to Sigmatic Sciences, a Sapio company. Sapio Elain is scientist-directed, with AI in the loop. Sigmatic is AI-directed, with the lab and scientist in the loop. Two modes, one partner network.
The thread running through it
What was clear from the Pistoia presenters and attendees was that the blockers aren’t ambition or budget. They are interoperability, data standards and the governance needed to make AI trustworthy at scale, something the Pistoia community is actively working on.