MoA-Prediction Predicting the mechanism of action of drugs using Machine Learning
Mechanism of Action Prediction was a data science competition hosted on Kaggle by Laboratory for Innovation Science at Harvard. The goal of the competition is advancing drug development through improvements to MoA prediction algorithms.
Data: Gene expression data and cell viability data, compound or control perturbations, time and dosage of drugs along with the targets. Evaluation Function: log loss. Problem Type: Multilabel Classification problem.
The code can be run in any of the following way: Run on local machine: The code can be downloaded or you can also clone the repository. The notebooks can be run using jupyter-notebook on your local machine. Make sure that you have all the required libraries installed.
The Pytorch-RankGauss-PCA-NN model got a score of 0.01861 on public dataset and 0.01630 on private dataset.
This was the very frist time I participated in a live Data Science Competition and I did learn a lot of things on the way. Thanks to the Kaggle communtiy. Although I was just starting out in Machine Learning, I did manage to reach among the top 30% in the competition. I plan to participate in more competitions in the future.