BENCHMARKING RESULTS FOR RiPP IDENTIFICATION

Method: Support Vector Machine (SVM)

Sensitivity	Specificity	Precision	MCC	AUC
0.94	0.90	0.90	0.85	0.97

BENCHMARKING RESULTS FOR RiPP CLASS PREDICTION

MultiClass SVM

Class	Sensitivity	Specificity	MCC
LanthippeptideB	0.89	0.98	0.88
LanthippeptideA	1.00	1.00	1.00
LanthipeptideC	0.67	0.99	0.76
Linaridin	0.67	0.99	0.77
Cyanobactin	0.93	0.97	0.88
Sactipeptide	0.00	1.00	0.00
Microcin	1.00	1.00	1.00
Lassopeptide	0.51	1.00	0.69
Bacterial Head-to-Tail Cyclized Peptide	1	1	1
Auto Inducing Peptide	0.75	1	0.86
ComX	1.00	1.00	1.00
Thiopeptide	1.00	0.99	0.96

Average Sensitivity is 0.78, Specificity is 0.99 and MCC is 0.82

BENCHMARKING RESULTS FOR LANTHIPEPTIDE CLEAVAGE PREDICTION

Total	Positive Set	Negative Set#	Sensitivity	Specificity	Precision	MCC
2314	52	2262	0.71	0.99	0.69	0.69

#Note: The 'cost factor' has been used while training the classifier to adjust the diffrences
in counts of positive and negative datasets

BENCHMARKING RESULTS FOR LANTHIPEPTIDE CROSSLINKS PREDICTION

Method: Random Forest (RF)

View Comparison of Predicted Crosslinks with Actual Structure for Lanthipeptides

Total	Positive Set	Negative Set#	Sensitivity	Specificity	Precision	MCC
1576	218	1358	0.72	0.95	0.73	0.68

#Note: The 'cost factor' has been used while training the classifier to adjust the diffrences
in counts of positive and negative datasets









Method: Supoort Vector Machine (SVM)

Total	Positive Set	Negative Set#	Sensitivity	Specificity	Precision	MCC
1576	218	1358	0.57	0.94	0.63	0.54

#Note: The 'cost factor' has been used while training the classifier to adjust the diffrences
in counts of positive and negative datasets

BENCHMARKING RESULTS FOR LASSOPEPTIDE CLEAVAGE & CROSSLINKS PREDICTION

Cleavage site Prediction AUC = 0.998
CrossLinks

Total sequences	Correct prediction in top rank	Correct prediction in top 2 rank
60	50(83.33%)	55(91.67%)

BENCHMARKING RESULTS FOR CYANOBACTINS CLEAVAGE & CROSSLINKS PREDICTION

1. Core peptide prediction
Two SVM Classifiers were used, one each for RSII and RSIII.

Model	AUC
RSII Predictor	0.96
RSIII	0.95

2. Prediction of heterocycle rings

Total Fragments	28
Positives (With Heterocycles)	21
Positives (Without Heterocycles)	7
AUC	1

BENCHMARKING RESULTS FOR THIOPEPTIDE CROSSLINK PREDICTION

Total Sequences	35
Correct Prediction	28
Incorrect Prediction	07
Accuracy	80%