🏥 From Protein Folding to Clinical Trials
Isomorphic Labs, the drug discovery company spun out of Google DeepMind in 2021 by Demis Hassabis, announced on May 16, 2026, that ISO-001—a small-molecule kinase inhibitor designed entirely through AI-driven computational methods—has entered Phase II clinical trials for non-small cell lung cancer (NSCLC) with EGFR exon 20 insertion mutations. This marks the first AI-designed drug to reach Phase II, a milestone that validates the thesis that protein structure prediction can materially accelerate drug development.
The development timeline is striking: AlphaFold 3 predicted 140,000 protein-ligand binding conformations across 32 kinase targets in 11 days of computation, identifying ISO-001's binding pocket within the first 6,000 predictions. From target identification to lead compound took 11 months, compared to the industry average of 4-6 years for traditional high-throughput screening approaches. Subsequent medicinal chemistry optimization (2 years) and Phase I safety trials (14 months) proceeded on conventional timelines, but the early discovery acceleration compressed the overall timeline by an estimated 40%.
đź“‹ AlphaFold 3's Expanded Capabilities
Google DeepMind released AlphaFold 3 in May 2024 with a major expansion beyond protein folding: the model now predicts structures and interactions for proteins, DNA, RNA, small molecules (ligands), and ions, as well as their complexes. This "all-atom" approach lets researchers model how drug candidates bind to disease targets directly, rather than inferring binding from static protein structures. The AlphaFold Server, a free web interface launched alongside the model, has processed over 200 million structure predictions and been used by 2.4 million researchers in 190 countries.
Isomorphic Labs built a proprietary drug design pipeline on top of AlphaFold 3 predictions, adding ligand property optimization (ADME-Tox, synthetic accessibility), molecular dynamics simulation at millisecond timescales, and generative chemistry models that propose novel compounds with desired binding profiles. The company's partnerships with Eli Lilly and Novartis—signed in 2024 and 2025—carry total milestone payment values of $4.2 billion, with Isomorphic retaining ownership of the underlying AI platform while pharma partners hold commercial rights to discovered compounds.
⚔️ AI Drug Discovery Competition
Isomorphic Labs operates in an increasingly crowded field. Recursion Pharmaceuticals, following its merger with Exscientia in 2024, has three AI-discovered compounds in Phase I trials. Insilico Medicine's INS018_055, an AI-designed anti-fibrotic, completed Phase IIa in idiopathic pulmonary fibrosis with positive biomarker signals.
Schrödinger's physics-based computational platform continues to supply candidates to Bristol Myers Squibb and Takeda. Eli Lilly CEO David Ricks has projected that 30% of Lilly's clinical pipeline will originate from AI-driven discovery by 2030.
Demis Hassabis has been explicit about Isomorphic's ambition: an AI-designed drug approved by regulators and available to patients by 2029. "We're not just trying to speed up drug discovery by 20 or 30 percent. We're trying to compress a 12-year pipeline into 4 years," Hassabis said at the 2026 World Science Festival. "If we succeed, the pharmaceutical industry fundamentally changes."