AI powered drug discovery
Google Deepmind spinoff Isomorphic Labs is close to raising $2bn in new funding to support their aim to accelerate drug discovery.
Isomorphic are using AI to solve problems that have been hard, or stubbon to solve in discovering new drugs.
Developing a single approved drug typically takes 12–15 years and can cost billions of dollars. Around 90% of drug candidates fail somewhere in their clinical trials, or even later when the investment has already been made.
A predecessor to Isomorphic, Googles Alphafold solved protein structure prediction, what biological molecules look like. Knowing the structure alone is not enough to create medicines.
Isomorphic Labs Drug Design Engine (IsoDDE) combines multiple predictive models into one platform, designed to run end‑to‑end drug discovery through computer based methods. It predicts protein–ligand structures more accurately, identifies novel binding pockets from sequence alone and redicts binding affinity. It has been shown to outperform some physics‑based methods at far lower cost.
Isomorphic are working with big Pharma to design its own therapeutics. With an initial focus on oncology, immunology and inflammation it is preparing its first AI‑designed candidates for clinical trials.
Their aim is shrink early discovery from years to months, dramatically reduce the failure rate before humans are dosed and unlock entire classes of diseases deemed “too hard” or “too risky”. A bold ambition.
It isn't all plain sailing though. Clinical biology is still unpredictable, regulatory approval remains slow and you can't simulate the long-term safety of the drug.
Isomorphic presents its work as accelerating and improving the process, rather than replacing it. What this does highlight is a strategic difference in how Google is applying AI, in scientific and niche areas, compared to Anthropic and OpenAI who building general‑purpose AI brains that work in many different domains.
SOURCE
