Table 1.
Fisher’s Exact Tests comparing deleteriousness predictions between binding and non-binding mutations
| # Binding Mutations | # Non-Binding Mutations | |||||
|---|---|---|---|---|---|---|
| Method | deleterious | tolerated | deleterious | tolerated | Odds Ratio | P-value |
| SIFT | 23094 | 9142 | 486 089 | 312 266 | 1.62 | ∼0 |
| PolyPhen2, HDIV | 21208 | 11801 | 416 954 | 400 778 | 1.73 | ∼0 |
| PolyPhen2, HVAR | 18574 | 14435 | 326 362 | 491 306 | 1.94 | ∼0 |
| MutationTaster | 6284 | 3868 | 129 866 | 94141 | 1.18 | 4e-15 |
| PROVEAN | 6411 | 3429 | 103 264 | 114 453 | 2.07 | 3e-262 |
| REVEL | 3230 | 6886 | 47823 | 175 674 | 1.72 | 3e-127 |
| MutPred | 6011 | 3156 | 92629 | 109 803 | 2.26 | 7e-305 |
Each distinct somatic mutation in the pan-cancer dataset is classified as either binding (i.e. falls into an InteracDome-inferred, confident putative binding position in at least one human protein) or non-binding. Corresponding deleteriousness scores for each of these mutations were retrieved, where available, from the Database for Nonsynonymous SNPs’ Functional Predictions (v3.5) (52); many mutations analyzed did not have corresponding deleteriousness scores for one or more predictors. Score thresholds to distinguish deleterious from tolerated mutations were set as recommended by each method or to ≥0.5 when not specified for REVEL and MutPred scores.