Jackson Burns
Jackson Burns
Embedded Files
CheMeleon Foundation Model
CheMeleon Foundation Model
Deep Learning Foundation Models for Low-Data Regimes from Classical Molecular Descriptors is out now in the Journal of Chemical Information and Modeling
This work descrives out efforts to create CheMeleon, a breakthrough foundation model for molecule property prediction spanning fields from drug discovery to materials science.
FASTSOLV
FASTSOLV
Solubility of small molecules in organic solvents is a critically important quantity for pharmaceutical development and formulation, process control, and fine chemical synthesis
a-priori estimation is challenging and available data is limited in accuracy
FASTSOLV is the best available model for organic solubility prediction
extrapolates to unseen solutes with 2-3x better accuracy than alternatives
solubility gradient versus temperature are incredibly accurate
inference times low enough to enable real-time high-throughput applications
Open access in Nature Communications
Source code and package on GitHub
Try for free in your browser: fastsolv.mit.edu
Page updated
Google Sites
Report abuse