Cheminformatics_projects

First project: predicting toxicity and solubility of molecules
include:
Deepcharm module and toxicity dataset
Model trained for toxicity using tox21_datasets and MultitaskClassifier
Model shows over-fitting as the test dataset scores are much less compared to the train-set. But model can predict toxicity of the testset molecules
Graph convolutional neural network model is trained on solubility dataset (delaney dataset from MoleculeNet)
Solubility values were predicted for new molecules, and we showed their structure using RDKit. Extra queries for substructures in sample molecules are also shown here.