Skip to main content

Table 1 Summary of ML models and features were used for training PVPs

From: Application of machine learning in bacteriophage research

No. Predictor Method Number dataset (TR/TS) Performance
1 ANN “ACC, protein isoelectric Points” + ANN 307 (307/NA) 85%
2 Naïve Bayes “ACC, DPC” + CFS + Naïve Bayes 401 (307/94) 79%
3 PVPred g-gap DPC + ANOVA+SVM 307 (307/NA) 85%
4 PhagePred g-gap DPC + ANOVA + Multinomial Naïve Bayes 307 (307/NA) 98%
5 PVP-SVM “AAC, ATC, CTD, DPC, PCP” + RF-based feature selection + SVM 401 (307/94) 87%
6 SVM-based g-gap DPC + “ANOVA, mRMR” + SVM 401 (307/94) 86%
7 Ensemble RF “CTD, bi-profile Bayes, PseAAC, PSSM” + Relief + RF 501 (253/248) 85%
8 Pred-BVP-Unb CT, SAAC, bi-PSSM+SVM 401 (307/94) 92%
9 PVPred-SCM DPC + SCM 401 (307/94) 77%
10 Meta-iPVP Probabilistic feature+SVM 626 (313/313) 82%
  1. SCM scoring card method, SVM support vector machine, AAC amino acid composition, ATC atomic composition, bi-PSSM bi-profile position specific scoring matrix, CTD chain-transition-distribution, CT composition and translation, DPC dipeptide composition, GDPC g-gap dipeptide composition, PCP physicochemical properties, SAAC split amino acid composition, TR training dataset, TS testing dataset