In Silico AI based solutions complement computational chemistry methods with higher speed and greater ability to learn the underlying patterns.
Accelerate the iterations in between Insilico & Invitro drug discovery processes using AI based models that enable more informed experimentation.
AI powered prediction models can predict binding affinities between a ligand and a protein and also estimate IC50 values using Extended Connectivity Fingerprints (ECFP).
Building discrete models for individual protein families enables higher accuracies and flexibilities for scientists to choose relevant models.
Using custom featurizers based on graph convolution operations on molecular graphs instead of conventional ECFP will improve accuracies of models built to predict ADME (Absorption-Distribution-Metabolism-Extraction) properties.
De-Novo Drug design can be pursed for ADME optimization by exploring ‘drug-like’ space using reinforcement learning techniques in conjunction with recurring patterns in known approved drug molecules.
AI models can be used to identify toxicophores in molecule leads using sub-structure molecular search over known databases of toxicophores, thus helping drug researchers filter out false leads.
Traditional methods such as Scaffold hopping can be more efficient by using AI based fragment replacement techniques while retaining the essential pharmacophores.