Indian scientists develop PathGennie to speed up drug discovery simulations

Scientists at the S. N. Bose National Centre for Basic Sciences (SNBNCBS), Kolkata, have developed a novel computational framework called PathGennie that can dramatically accelerate the simulation of rare molecular events, a long-standing challenge in computer-aided drug discovery (CADD).

Published in the Journal of Chemical Theory and Computation , PathGennie is an open-source software that predicts how drug molecules unbind from their target proteins without relying on artificial forces or elevated temperatures that often distort results in conventional methods.

In pharmaceutical research, a key parameter in drug effectiveness is the residence time the duration a drug remains bound to its target protein. While binding affinity is important, residence time often provides a better indication of how well a drug will work in the body. However, simulating the unbinding process is notoriously difficult because it involves “rare events” that occur over milliseconds to seconds, far beyond the reach of standard molecular dynamics (MD) simulations, even on powerful supercomputers.

To overcome this, researchers traditionally force such events using biasing forces or high temperatures, which can alter the natural physics of molecular interactions and lead to inaccurate predictions. PathGennie avoids this pitfall by adopting a fundamentally different strategy.

Developed by a team led by Prof. Suman Chakrabarty , along with Dibyendu Maity and Shaheerah Shahid , PathGennie employs a direction-guided adaptive sampling approach that mimics natural selection at the molecular level. Instead of forcing molecules to move, the algorithm launches large swarms of ultrashort, unbiased MD trajectories each lasting only a few femtoseconds and selectively extends only those trajectories that show progress toward a defined end state.

This “survival of the fittest” strategy for molecular trajectories allows the method to bypass long waiting times associated with rare events, while preserving the true kinetic pathways of molecular transitions. The approach can operate in a wide range of collective variables, including high-dimensional or machine-learned representations, making it both flexible and broadly applicable.

In proof-of-concept studies, PathGennie successfully uncovered multiple competing unbinding pathways in complex systems. It rapidly mapped how a benzene molecule escapes from the deep binding pocket of the T4 lysozyme enzyme and identified three distinct dissociation pathways for the anti-cancer drug imatinib (Gleevec) as it unbinds from the Abl kinase. Notably, the method recovered all previously reported pathways using only a few iterations and without applying any external steering forces, closely matching results from earlier biased simulations and experimental studies.

Beyond drug discovery, the researchers say PathGennie can be applied to a wide range of problems involving rare events, including chemical reactions, catalytic processes, phase transitions and self-assembly phenomena. Its compatibility with machine-learning techniques further enhances its potential for integration into modern simulation workflows.

By making the software freely available, the team hopes PathGennie will lower barriers for researchers worldwide and contribute to faster, more accurate modelling of complex molecular processes, with significant implications for drug development and materials science.

PathGennie is a powerful computational method that accelerates the simulation of rare molecular events by identifying realistic transition pathways without using artificial forces or high temperatures, thereby preserving true molecular physics. It efficiently reveals multiple drug unbinding routes and is flexible enough to work with machine-learned variables, making it useful beyond drug discovery. However, it requires a clearly defined end state, depends on the choice of collective variables, and complements rather than replaces traditional long-timescale simulations.