The IntFold Team Leon Qiao Wayne Bai He Yan Gary Liu Nova Xi Xiang Zhang

Abstract

We introduce IntFold, a controllable foundation model for both general andspecialized biomolecular structure prediction. IntFold demonstrates predictiveaccuracy comparable to the state-of-the-art AlphaFold3, while utilizing asuperior customized attention kernel. Beyond standard structure prediction,IntFold can be adapted to predict allosteric states, constrained structures,and binding affinity through the use of individual adapters. Furthermore, weintroduce a novel confidence head to estimate docking quality, offering a morenuanced assessment for challenging targets such as antibody-antigen complexes.Finally, we share insights gained during the training process of thiscomputationally intensive model.