Computational Drug Discovery

4-None

Foundation models, large neural networks pretrained on enormous unlabeled corpora, have become the dominant computational tool in biology over the last five years. Protein language models such as ESM-2 and ESM-3 predict 3D structure & design new proteins from sequence alone. Genome-scale models such as Evo 2 predict pathogenicity for clinical variants without supervised fine-tuning. Single-cell models such as Geneformer and Universal Cell Embeddings represent millions of cells across species in shared coordinate spaces. AlphaFold 3 and Boltz-2 unify protein, nucleic acid, and small-molecule structure prediction in single architectures. Phenotypic-profiling models trained on Cell Painting microscopy have entered drug discovery pipelines.