Key features selected by IDEPI for each of the example problems.

Notation: T297K means that K is found in position 297 (HBX2 coordinates, T is the residue found in HXB2); PNGS (T188) – a potential N-linked glycosylation site with N at HXB2 coordinate 188; PNGS (N234+N276) – a pair of potential N-linked glycosylation site with N at HXB2 coordinates N234 and N276; [] – a deletion relative to HXB2. The ranking of the features is based on what order they were added to the model, and averaged over cross-validation replicates. For datasets with little signal (e.g. 10E8 bNab, 8ANC131 bNab), there was considerable variation in feature ranks among CV replicates, hence the best ranking feature has a median rank worse than 1. The values in the MCC column are for the models with the corresponding number of features (e.g. the MCC of a 2-feature model for V3 tropism in 0.67).
Key features selected by IDEPI for each of the example problems.