Bridging CGR Representations and Language Models for Reaction Property Prediction

LLMs predict material properties based on atom types and not atom positions

Notation matters: cross-representation inconsistency in chemistry language models and its mechanistic origins

Large language models (LLMs) are increasingly applied to molecular property prediction, but whether they encode notation-agnostic molecular representations or exploit surface-level string patterns re- mains untested. We present a dual-phase...

TopoFormer: A Topological-Geometric Neural Network for Accurate Protein Solubility Prediction

Protein solubility is a key consideration for the production of proteins at industrial scales. Due to the complex and nonlinear nature of the physical processes involved in protein solubility, existing...

SurGBSA: Learning Representations From Molecular Dynamics Simulations

Protein-ligand interactions are inherently dynamic, yet most neural network methods for structure-based drug design rely on static structural representations. Atomistic molecular dynamics (MD) simulations provide a tool to model the...

Swarms of Large Language Model Agents for Protein Sequence Design with Experimental Validation

Designing proteins de novo with tailored structural, physicochemical, and functional properties remains a grand challenge due to the vastness of sequence space and the complex coupling between sequence, structure, and...

Accelerated screening of cellulose additives via supervised latent representations

We combined chemical descriptors with atomistic simulations to construct a physics-informed latent representation that predicts the effects of molecular and polymeric additives on regenerated cellulose films.