The accuracy (%) of multi-class MIL by using ESM-1b and k-mer features on tiny host datasets.

Comparison of mean accuracy and standard deviation between ESM-1b and k-mer feature sets: ESM-1b, AA_2, PC_3 and DNA_5. For each feature, training multi-class classification models on 22 prokaryotic hosts and 36 eukaryotic hosts based on 5-fold cross-validation, then mean and standard deviation of accuracy is obtained by testing the trained model on host datasets that are associated with fewer viruses, ranging from 5 to 30 (tiny host datasets). (XLSX)