Domino Data Lab’s VP of Life Sciences Strategy has made five key predicitons
Domino Data Lab’s vice president of Life Sciences Strategy, Chris McSpiritt, has made five predictions on how he believes AI will play out in 2026.
Prediction One
By the end of 2026, around one-third of discovery labs will operate closed-loop DMTA systems, advancing on pilot programs that have shown five times faster design-to-synthesis cycles.
These platforms can connect generative chemistry models, robotic synthesis, and automated assays that operate overnight and retrain on new data by morning.
Human researchers can set priorities and constraints, while loops execute and learn without human intervention, compressing hit-to-lead cycles from the current multi-month average to under four weeks.
Prediction Two
Throughout next year, digital ADME-Tox simulations could replace around 30% of preclinical assays in pharma pipelines, reducing per-compound evaluation costs by about 40% and minimising animal use across pre-IND packages.
High-fidelity models combining molecular structures, omics data, and prior results can filter out low-value compounds prior to being sent to the lab.
Prediction Three
By December 2026, over half of the top 50 biopharmas will utilise AI evaluators aligned with ICH M11 and CDISC standards to model and score clinical trial protocols before first-patient-in.
These evaluators will be able to test inclusion criteria, visit schedules, and endpoints against diversity, cost, statistical power, and feasibility, enabling teams to cut mid-study amendments by 20%, and the “design once, simulate many times” approach is thought to become the industry standard for greater efficiency and decision-making.
Prediction Four
Throughout 2026, regulators will start to expect traceable, versioned LLM copilots in statistical computing environments (SCE) as part of a risk-based validation framework outlined in the FDA’s 2025 AI guidance and the EU AI Act.
In response, quality and compliance teams will begin auditing these copilots like code, verifying data lineage, model parameters, and human sign-offs under 21 CFR Part 11 controls. These copilots will become validated components of the analytical stack.
Prediction Five
Throughout 2026, regulators will closely review AI feature stores, model registries, and lineage systems during standard GxP audits. Model dossiers with training data summaries, validation reports, and change control plans will become a requirement for every protocol and clinical study report submitted for review.
This will advance AI governance to a necessary operational requirement. Organisations that establish continuous validation across all analytics will be able to earn regulatory trust and first-mover advantages.
