Current AI companies are suffering from insufficient data.
Prediction of drug potency (DP) requires an accurate AI model and big data to train it. Currently, however, very insufficient experiment data for DP exists and therefore AI engines based on it provide only partial and doubtful information for a novel drug candidate.
We generate data itself of experiment-level via physics-based method.
Most AI-based drug development companies are using only experiment data that are small in number to faithfully train AI engines. On the contrary, we build a reliable big data using Molecular Dynamics calculation which can precisely simulate real structures and dynamics of a protein-ligand complex.
Drug screening platform with a bottom-up approach
Until very recently, AI models for drug discovery have adopted top-down approaches, i.e., statistical inference methods that are consequently not structure-based. Our machine learning (ML) models are based on the structure-based data of accurate binding affinities between proteins and ligands. Furthermore, we incorporate physics-based domain knowledge of atoms into our ML model which hence can offer a more reliable prediction.