ChemAI 2025

Revolutionizing Materials R&D with Combined Physics-Based and Machine-Learning Approaches

Speaker:
Anand Chandrasekaran, Senior Principal Scientist at Schrödinger

Abstract:
Advances in materials discovery increasingly rely on merging the predictive power of physics-based simulation with the speed and adaptability of machine learning. At Schrödinger, we integrate molecular dynamics (MD), quantum mechanics, and data-driven models to accelerate property prediction and design.

In collaboration with SABIC, machine-learning models augmented with MD simulations accurately predicted polymer glass-transition temperatures, dielectric constants, and refractive indices, guiding the selection of next-generation polycarbonates. With Panasonic, large-scale MD and reinforcement learning were combined to design and experimentally validate ultra–low-viscosity solvents for advanced electrolytes.

Extending these principles, our Formulation ML solution predicts the properties of complex mixtures by linking molecular structure, composition, and simulation-derived descriptors to target physical properties. Together, these efforts show how integrating physics-based insight with machine learning accelerates innovation, improves interpretability, and delivers experimentally validated materials far faster than traditional R&D.