Drug Discovery Technologies - overview


Congratulations to IBMer Jannis Born (IBM Zurich Research Labs) for winning the Roche FXH2019 Scientific Excellence Award for the PaccMann deep learning tool for explaining cancer drug sensitivity!  See details in News tab.

IBM Research leverages deep and long-standing expertise in the modeling and simulation of protein-ligand complexes together with latest AI under the leadership of a 20 year pharmaceutical industry veteran to develop cutting-edge atomistic technologies to support lead finding and lead optimization. A number of research threads are in progress, starting with deep learning representation for target activity prediction using ligand information only (QSAR) or additionally protein structure information (docking). Attention mechanism models have been integrated to link specific small molecule substructues with predicted activity, whether desired on-target or undesired off-target. In addition, we are developing generative AI methods to create molecules with specified profiles, rather than simply scoring input molecules. A long term goal is to integrate protein dynamic information into the deep learning and generative approaches but in a higher throughput manner relative to current approaches where each new protein-ligand system requires a new simulation.