Cyclica: predicting drug binding sites

Inspiration

Predicting where the drug binding site is located on a protein is a costly task.

What it does

Predicting drug binding sites on AlphaFold2-predicted proteins

How we built it

DecisionTree, GradientBoost

Challenges we ran into

1. Not understanding the variable's meaning
2. Preprocessing is a tedious process that takes a long time
3. EDA, preprocessing, and modeling are intertwined
4. Failed to agree on the same model to use at the end of the context, check out submission sites, and requirements on the last day

Accomplishments that we're proud of

Mastery of basic Pandas, and Numpy functions. Familiarity with Sk-learn

What we learned

1. Consulting with students or experts in the fields creates inspiration to tackle the challenge
2. Consistent communication is important to project outcome and quality
3. Focus on the big picture, don't be swayed by small details

Built With

• scikit-learn
• tensorflow
• xgboost