Atomwise's ML superplatform employs deep learning models, particularly convolutional neural networks applied to 3D molecular structures, to predict binding affinities between small molecules and protein targets. The system is capable of virtually screening billions to trillions of compounds in silico, far exceeding the throughput of traditional high-throughput screening (HTS) methods. By learning from large datasets of protein–ligand interactions, the platform predicts drug-like properties, selectivity, and potency to prioritize synthesis candidates. This approach substantially compresses the hit-identification phase of the drug discovery pipeline and enables structure-based lead optimization at scale.