Overcoming PAINS in AI Drug Discovery
Papers
Navya Ramesh
Model Medicines GALILEO™ Platform’s AI Graph Mining Approach Classifies Pan Assay INterference compoundS (PAINS) with Best-in-Class Model Performance Metrics
Navya Ramesh
Fellow, Machine Learning Engineer
In our latest research, Model Medicines' Graph-Mining approach outperformed other PAINS classification models for AI drug discovery.
PAINS stands for Pan Assay INterference compoundS in AI drug discovery. It refers to compounds that have a common substructural motif that encodes for an increased chance of any member registering as a hit, leading to false positive and false negative "discoveries" that waste time and resources.
Our best-in-class approach significantly outperformed traditional models like Structure Filter, ML-RF, and GNN, as well as literature reported SVM, ML-RF DNN, and Graph Mining models.
This paper demonstrates how we're optimizing drug discovery and reducing the number of PAINS compounds in both training and inference data sets. Our proprietary AI Graph-Mining PAINS classification engine improves the efficacy, efficiency, and ROI of our GALILEO™ AI Drug Discovery platform.
