During our lifetimes, we are exposed to different chemicals and some of which may be toxic and harmful to our health. It is ideal and promising that a time saving and human-free toxicity prediction methods is developed for human or animals involved toxicity tests are quite time-consuming and dangerous. Deep neural networks, in this case, are ideal to conquer this problem with large amount of data. GCN, with an accuracy of 88.798% are used to solve the problem.
The data set is from the kaggle competition Drug Molecular Toxicity Prediction.
| AUC of Training set and Validation set | AUC of Test set |
|---|---|
| 87.964% | 88.798% |