Software
Tools and software developed by our research group
ViSNet-PIMA & AI²BMD-PIMA
ViSNet-PIMA introduces a Physics-Informed Multipole Aggregator to capture non-local interactions in biomolecules. It outperforms SOTA models on MD22/AIMD-Chig datasets and, via the AI²BMD-PIMA framework, reduces protein simulation errors by over 50%, bridging the gap between ML efficiency and ab initio precision.
AI²BMD
AI²BMD is a machine learning framework for efficiently generating ab initio quality molecular dynamics trajectories. This updated version introduces a robust geometry optimization module, allowing for seamless structural refinement before dynamic simulations. Furthermore, the enhanced Apptainer implementation now provides broader hardware compatibility, supporting a wider range of GPU models to ensure stable, high-performance deployment across diverse computing clusters.