Atomwise

Atomwise applies deep learning to structure-based drug discovery, using its AtomNet convolutional neural network to predict how small molecules bind to protein targets. Instead of running physical assays on every candidate, pharma and biotech partners can screen billions of virtual compounds computationally and narrow the field down to a manageable shortlist for lab testing and further development. AtomNet was trained on large sets of known protein-ligand structures, which lets it evaluate binding affinity for targets even when little or no prior chemistry data exists for them.

The company works mainly through R&D collaborations with pharmaceutical and biotech partners rather than selling a self-serve software product. It has run projects with companies including Sanofi, Eli Lilly, Bayer, and Hansoh, spanning targets in oncology, inflammatory disease, and neurodegenerative conditions, and has also advanced some of its own internal programs, including a TYK2 inhibitor, toward clinical testing. Deals typically combine upfront access fees with milestone payments tied to how far a discovered compound progresses.

What differentiates Atomwise from many AI drug discovery entrants is its head start: AtomNet was one of the first deep learning models applied to structure-based virtual screening, giving the company years of accumulated project data across hundreds of targets and partnerships. That track record, rather than a public-facing tool, is the core of its pitch to pharma partners.