Isomorphic Labs and Insilico Medicine advance AI drug discovery

Isomorphic Labs expects to begin first-in-human testing by the end of 2026. The company uses its IsoDDE engine to predict protein-ligand interactions and identify new binding pockets from amino acid sequences alone, which doubled the performance of AlphaFold 3 on difficult protein-ligand structures in a February 2026 report. This capability allows the platform to predict binding affinities better than physics-based methods. Novartis expanded its January 2024 collaboration with Isomorphic in February 2025 to include three additional research programs. This Novartis agreement includes up to $1.2 billion in milestones and royalties. Isomorphic focuses on oncology and immune-mediated disorders. The company uses IsoDDE to model protein-ligand complexes and antibody-antigen interactions. In a February 2026 report, Isomorphic stated that IsoDDE provides 2.3 times better high-fidelity prediction of antibody-antigen interactions than AlphaFold 3. While AlphaFold 3 improves accuracy in modeling interactions, it struggles with large conformational changes and shows conformational biases that limit its reliability for certain targets. Isomorphic Labs raised $2.1 billion in a May 2026 funding round led by Thrive Capital to scale its drug design engine. This follows a $600 million Series A round in April 2025. The company also maintains research collaborations with Eli Lilly, which include up to $1.7 billion in performance-based milestones.

Insilico Medicine accelerates target-to-candidate workflows

Insilico Medicine signed a $2.5 billion research and development collaboration with SK Biopharmaceuticals at the BIO 2026 convention. This alliance targets neuroimmune disorders such as neuroinflammatory and neurodegenerative conditions. The Pharma.AI platform reduces target-to-preclinical candidate timelines to 12 to 18 months. Traditional drug discovery processes take up to 4 years to reach the same stage. Insilico has nominated 31 preclinical candidates since 2021. Thirteen of these assets received IND approval or clearance. Insilico is eligible to receive up to $18 million in upfront and near-term milestone payments from the SK deal. The Chemistry42 platform combines generative models with physics-based methods for hit identification and lead optimization. Researchers use it to generate novel molecules and optimize them against multiple properties. PandaOmics uses the PandaClaw agent to automate detailed analytical workflows, including planning and report generation. This tool assists in target evaluation and target-disease evaluation. Insilico’s "AI trains AI" approach uses foundation models and scientific AI agents to create a unified paradigm for R&D. Insilico also uses the MMAI Gym for training and benchmarking foundation models, which can produce a 10x performance increase over baselines. In a study published in Chemical Science, researchers used AlphaFold-predicted structures and the Chemistry42 platform to discover a novel liver cancer drug candidate in 30 days.

The industry moves toward clinical validation

The pharmaceutical industry reached a milestone in 2026 as 200 AI-discovered drugs entered active development. This pipeline includes 94 candidates in Phase I trials, 56 in Phase II, and 15 in Phase III. AI-discovered drugs succeed at nearly double the rate of traditional compounds. A typical drug discovery and development process takes 10.5 years to reach completion. This average includes 2.3 years for Phase I, 3.6 years for Phase II, 3.3 years for Phase III, and 1.3 years for regulatory filing. The cost to discover and develop a single FDA-approved drug is approximately $2.87 billion. The FDA released draft guidance in 2025 regarding the use of AI to support regulatory decision-making via a risk-based framework. This guidance requires model transparency and validation at critical decision points. Will these AI-designed molecules behave in humans the same way they do in computational simulations? You should monitor the Phase II results for Insilico’s idiopathic pulmonary fibrosis candidate expected in mid-2026. The industry is moving toward a model where AI handles the in silico design and prediction while partners like Novartis or Lilly manage experimental validation.