A comparison of iLoc-Euk with Euk-mPLoc 2.0[40] using the jackknife cross-validation test on the benchmark dataset taken from the Online Supporting Information S1 of [40].

The dataset contains 7,766 different eukaryotic protein sequences covering 22 location sites where none of the proteins included has pairwise sequence identity to any other in a same location.aThe predictor from [40].bThe predictor proposed in this paper.cNote that instead of 7,766 (the number of total different proteins), here we use 8,897 (the number of total different virtual proteins) for the denominator. This is because some proteins may have two or more location sites. As for the definition of “virtual protein”, see Eqs.2–3 of [40] and the relevant explanation there.