VIRsiRNApred: a web server for predicting inhibition efficacy of siRNAs targeting human viruses

Table 1 Ten-fold cross validation performance of predictive models on viral siRNA dataset of 1380 sequences (T ¹³⁸⁰ ) using SVM, ANN, KNN and REP Tree machine learning techniques

Predictive model no.	siRNA features		No. of siRNA features	Pearson correlation coefficient* on training (T¹³⁸⁰) dataset# during 10-fold cross validation
Predictive model no.	siRNA features		No. of siRNA features	SVM	ANN	KNN	REP Tree
1		Mononucleotide frequency	4	0.19	0.10	0.11	0.10
2		Dinucleotide frequency	16	0.32	0.29	0.29	0.29
3		Trinucleotide frequency	64	0.42	0.28	0.30	0.28
4		Tetranucleotide frequency	256	0.43	0.28	0.30	0.30
5		Pentanucleotide frequency	1024	0.46	0.29	0.30	0.30
6		Binary	76	0.19	0.10	0.11	0.11
7		Thermodynamic features	21	0.26	0.22	0.21	0.20
8		Secondary structure	28	0.07	0.04	0.04	0.04
9		1 + 2 + 3 + 4 + 5	1364	0.48	0.30	0.31	0.31
10		6 + 9	1440	0.50	0.36	0.41	0.32
11		6 + 7 + 9	1461	0.55	0.46	0.48	0.45
12		6 + 7 + 8 + 9	1489	0.53	0.42	0.44	0.42

*Pearson Correlation Coefficient (PCC) is the correlation between experimental and predicted viral siRNA efficacy.
#T¹³⁸⁰ is the training dataset of experimental viral siRNA. Predictive Models 1-8 were developed on individual siRNA features while models 9-12 were based on hybrid siRNA features.

ISSN: 1479-5876