Kinase Selectivity Prediction with MACHINE LEARNING

Inspiration

Challenge 1 spoke to us.

What it does

It predicts kinase selectivity based on a dataset.

How we built it

Using XGBoost classifier.

Challenges we ran into

How to pre-process data cells that have a value of 10001

Testing different models

Tuning hyper parameters

Finding the best features that correlate to sensitivity

Accomplishments that we're proud of

Actually getting results

What we learned

What is a kinase?

What is machine learning?

Thank you the PharmaHacks teams for this opportunity to learn!

What's next

MSA to identify commonalities among Kinase families (active sites, binding sites, etc.), then use the embeddings of those common sequences to help our models

Built With

• colab
• google-notebook
• hugging-face
• jupyter
• matplotlib
• python
• pytorch
• rdkit
• scikit
• scikit-learn
• transformers
• xgboost