Artificial intelligence department

AI enabled HTVS

Our groundbreaking HTVS platform enables precise screening for any target using the expansive ENAMINE Real database, boasting billions of compounds. Our technology addresses the crucial step of target validation, and with the aid of advanced molecular modeling algorithms, we further support the elucidation of the mechanism of action of known binders. This knowledge equips us to dock an initial set of molecules and then train a graph-based model. This AI-driven model is capable of predicting against the billion-size compound collection in less than a day, accelerating your drug discovery journey.

De novo ligand generation

Our platform uniquely leverages Natural Language Processing (NLP) and graph-based learning algorithms that learn from SMILES notation and comprehensive molecular topology and properties data. Our active learning framework hones the synergy between in silico affinity evaluations and deep learning generative models. These models are adeptly engineered to create drug-like compounds, providing a fertile ground for therapeutic innovations.

ADMET prediction models

We offer a comprehensive suite of models, each rigorously trained on experimentally validated data and vetted against cutting-edge approaches. Our models provide reliable predictions for 21 ADMET endpoints, bolstering your R&D capabilities. To enhance transparency and utility, we provide confidence intervals alongside each property prediction and delve into the molecular topology responsible for the outcome.

De novo protein generation

Our protein language model has been trained on an extensive data set encompassing more than 350 protein superfamilies. This equips us to perform crucial downstream tasks such as conditional protein generation and protein-protein interaction prediction. We have created a robust protein embedding space that can be efficiently sampled to generate proteins from a family of your choice, all through fine-tuning to experimentally validated sets provided.

Protein-Protein Interaction Prediction

We extend the utility of our protein generation module to predict protein-protein interactions, a vital aspect of biological function and drug design. Our technology leverages vast training data and powerful machine learning algorithms to predict and analyze complex protein interactions. This allows us to anticipate potential binding sites, interaction dynamics, and therapeutic implications, providing you with an invaluable tool for drug discovery and development.