# Decrypting cryptic pockets with physics-based simulations and artificial intelligence

## TL;DR

This review discusses how physics-based simulations and AI help identify hidden protein pockets for drug discovery.

## Contribution

The paper reviews hybrid computational strategies combining physics and AI for discovering cryptic pockets.

## Key findings

- Physics-based simulations improve detection of transient protein pockets.
- AI-driven models enhance the functional interpretation of cryptic pockets.
- Hybrid methods offer better accuracy in identifying druggable sites.

## Abstract

Cryptic pockets are promising targets for drug discovery that greatly expand the druggable proteome. In particular, they can provide opportunities to target proteins previously thought to be “undruggable” due to a lack of pockets in structures of the ground state. However, their transient and hidden nature renders them difficult to detect through conventional experimental screening methods. Recent advances in computational methodologies and resources have greatly enhanced our ability to identify and characterize such elusive pockets. This review highlights key developments in computational approaches, including physics-based molecular dynamics simulations, artificial intelligence–driven models, and hybrid strategies that integrate both to enhance cryptic pocket discovery and functional interpretation.