Machine learning models to prioritize optimal parameters of predicted ADME and Toxicity data
Machine learning models to be generated to prioritize hits generated from virtual screening of large datasets. Any decision on prioritizing hits, leads is always based on multiple criteria, like aqueous solubility, docking score, ADME parameters or synthetic feasibility among others. Any statistical or machine learning algorithms could be used for the decision making process by considering optimal balance of ADME, Toxicity and virtual screening properties. The ML models should help to make decisions on drug candidates considering all properties required for a chemical compound to be called as 'drug-like' molecule focusing medicinal chemistry parameters for further assay studies. The parameters for checkpoints are: A) Decision Model generation - i) pharmacokinetic properties (from published data) models to generate threshold range and optimal balance of properties to be considered for candidate selection criteria. B) Deriving medicinal chemistry properties as well as the synthetic feasibility and accessibility for novel molecules. C) Model quality - i) parameters to define the model quality, applicability domain, reproducibility and reliability. Submission requirements: i) Model with a sample dataset where predicted ADME, Toxicity, docking score, free binding energy in kJ/mol are considered. Details of training, test and validation sets to be shared. ii) Model deployment for validating the predictions from Track 1 iii) Detailed report on the model and tutorial user guide. The above parameters could be calculated using any open tools, models and codes could be shared on a versioning platform for reproducibility