Insilico Medicine and Memorial Sloan Kettering Join to Apply AI Target Discovery to Gastri

Gastric and gastroesophageal junction cancers are among the most treatment-resistant solid tumors in oncology. Despite advances in HER2-targeted therapy and, more recently, checkpoint immunotherapy, the majority of patients diagnosed with advanced disease do not achieve durable responses. The biology of these cancers is heterogeneous - different molecular subtypes, different driver mutations, different immune microenvironments - which means finding treatments that work across the patient population requires identifying targets that address that heterogeneity rather than applying a single approach to everyone.

Insilico Medicine, a Hong Kong-listed generative AI drug discovery company, and Memorial Sloan Kettering Cancer Center announced a collaborative research agreement in February 2026 aimed at accelerating target identification for gastroesophageal cancers. The partnership pairs Insilico's AI-driven target discovery infrastructure with MSK's clinical and molecular data resources under the scientific direction of Yelena Y. Janjigian, MD, who leads MSK's GI Oncology program and whose group has been responsible for several practice-changing advances in gastric cancer treatment.

What PandaOmics Does

Insilico's PandaOmics platform is a multi-modal AI system designed to identify druggable biological targets by integrating diverse data types: genomic, proteomic, transcriptomic, and biomedical text data. The platform applies more than 20 proprietary AI and bioinformatics models to systematically rank candidate targets based on their biological plausibility as disease drivers and their potential tractability as drug targets.

The platform's design addresses a specific bottleneck in drug discovery: the gap between generating large amounts of multi-omics data and identifying which signals in that data correspond to targets worth pursuing. Human analysts can identify obvious targets from data, but the volume and complexity of modern multi-omic datasets exceeds what unaided analysis can efficiently process.

MSK's Data Contribution

MSK's contribution to the collaboration is its patient data: high-quality genomic, proteomic, and transcriptomic data from deeply annotated clinical cohorts covering diverse gastroesophageal cancer subtypes. MSK treats a large volume of GI malignancies and has built systematic infrastructure for collecting and organizing molecular data alongside detailed clinical information about diagnosis, treatment, and outcomes.

That combination - molecular data with matched clinical outcomes - is what makes the analysis potentially useful for identifying targets with translational relevance. A biological target that appears frequently in sequencing data but does not correlate with clinical behavior is less interesting than one that distinguishes patients who respond from those who do not, or that is specifically elevated in the subtypes most likely to cause death.

Dr. Janjigian described the motivation: "GEC patients need new breakthroughs, and those breakthroughs must come from a deeper understanding of each patient's unique disease biology. By combining patient-level clinical and molecular data with transformative AI tools, we can accelerate the discovery of clinically meaningful targets."

Where the Project Currently Stands

The collaboration is in its early phases. Current work centers on data gathering, quality control, and integration - establishing the technical infrastructure that allows MSK's datasets to be analyzed systematically within the PandaOmics framework. Subsequent phases are planned to involve AI-driven hypothesis generation, target ranking, and biological validation of prioritized candidates.

Alex Zhavoronkov, PhD, Insilico's founder and CEO, positioned the collaboration as a test of whether the company's AI platform can add value in a specific clinical challenge: "Gastroesophageal cancers remain among the most challenging solid tumors. By integrating MSK's exceptional clinical data resources with our target discovery technologies, we aim to identify meaningful biological insights and accelerate the development of new therapeutic options."

The Broader Context

Insilico reports that its AI-assisted approach has allowed it to nominate 20 preclinical drug candidates between 2021 and 2024 at an average timeline of 12 to 18 months per program, with 60 to 200 compounds synthesized and tested in each program. Traditional early-stage drug discovery typically requires 4 to 5 years to reach the same stage.

Whether that speed translates to clinical success remains to be demonstrated. Insilico's lead compound, a drug for idiopathic pulmonary fibrosis discovered using its AI platform, has reached Phase 2 clinical trials. The gastroesophageal cancer collaboration represents an application of the same AI-discovery approach to a disease area where the company does not yet have a clinical candidate.

Target identification is the first step in a long pipeline. Even if the collaboration successfully identifies novel druggable targets in gastroesophageal cancer, the path from target to approved therapy involves medicinal chemistry, safety testing, and multi-phase clinical trials that typically span a decade. The value of the collaboration will depend on whether the targets it identifies are genuinely new, biologically validated, and clinically meaningful.