Taxonomic bias in AMP prediction of invertebrate peptides.

However, we found no strong evidence for that, since neither the number nor the proportion of common amino acid sequences in our and the used tools’ data sets showed significant association with F 1 scores, although the proportion of common sequences showed a weakly significant association with MCC (see Supplementary Material Figs.

also in 1,387 other papers

Unexpectedly, there was a weak trend that when a query taxon was present in the training set of a tool, then F 1 and MCC scores tended to be slightly smaller (see Supplementary Material Fig.

also in 1,404 other papers