**Scheme 3: Lyssiotis collaboration | Testing to see if partitioning dataset into responder/nonresponder results in better accuracy / precision / recall / F1-score / MCC**

[ScottCampit]

• Train random forest classifier with all features for Lyssiotis dataset 2 (Input: cancer cell line for responder/nonresponder Output: patient derived cell line for responder/nonresponder)
• Train random forest classifier with only 1 subset of features for Lyssiotis dataset 2
• LOOCV