The Matthews correlation coefficient (MCC) evaluates binary classifiers using all four confusion-matrix entries: true positives, true negatives, false positives, and false negatives. It is especially useful when class frequencies are uneven. section 3.4.4
Interpret MCC through its range and compare it against relevant baselines:
- +1 indicates perfect classification, 0 represents random classification, and −1 indicates complete disagreement between predictions and actual classes. section 3.4.4
- Compute MCC on a held-out test set and compare it with baseline or published models rather than treating the score in isolation. section 3.4.4
hERG blocker prediction
A model evaluated on the hERG blockers test set achieved an MCC of 0.452, compared with 0.576 for a published model using more complex feature generation and modeling. section 3.4.4

