Computational drug discovery platforms propose small molecule designs 1000x faster than wet labs return experimental data. This is quickly becoming the bottleneck in modern drug discovery. We describe a platform designed to close that gap: bring your own model, and we return complete, decision-ready data every cycle, ready for model retraining. The next, data-informed turn of the loop is available on demand.

The platform, built on Ginkgo's scientific expertise and nearly two decades of lab-technology innovation, combines cutting-edge perturbation screening (DRUG-seq, Cell Painting, Metabolomics) for target validation or hit identification with ADME and developability assays (SMOL and Ab) built for scale. Together these form the foundation that lets teams run iterative lab-in-the-loop optimization or model enhancement.

Run properly, the platform becomes a complete Bayesian optimization loop. Your initial model holds the priors, your algorithmic experimental design is the search strategy, and each round of new data retrains the model and updates those priors. Applied to a well-defined experimental space, the self-learning loop is deterministic. It either stalls out, giving a data-driven stop condition, or converges toward the fitness function your team defines. Success becomes a matter of mathematics, not serendipity.