MUMAL: Multivariate analysis in shotgun proteomics using machine learning techniques

BMC Genomics

Table 1 Comparison between NN and SVM (LibSVM and SMO)

Dataset	NN	LibSVM	SMO
S1_NPH_CH2	318	158	174
S1_NPH_CH3	398	138	95
S1_PH_CH2	132	88	87
S1_PH_CH3	210	48	111
S2_NPH_CH2	72	37	50
S2_NPH_CH3	88	34	40
S2_PH_CH2	176	71	120
S2_PH_CH3	236	154	139
S3_NPH_CH2	72	-	-
S3_PH_CH2	487	231	413
S3_PH_CH3	338	147	295

The values indicate the number of PSMs that the learning method could retrieve when considering a 1% FDR. The NN values were significantly better. The dashes indicate that the algorithm could not find a set of hits with 1% FDR, i.e., there is no point in the ROC curve corresponding to such an error rate.

ISSN: 1471-2164