False discovery rates in spectral identification

BMC Bioinformatics

Table 9 Comparison between searches with differently normalized scoring functions

Search#	Spectra	Database	PMTol	Score	EmpiricalFDR fixed			FactFDR fixed
					N _target	FactFDR(%)\	p-value(%)	N _target	EmpiricalFDR(%)
I-1	ISB-02	ISB	2.5 Da	SpecProb¹	2329	5.8	10.9	2279	4.4
I-19	ISB-02	ISB	2.5 Da	MSGFRaw²	2079	4.6	36.5	2079	4.6
I-5	ISB-02+AB-TC	ISB+Yeast	2.5 Da	SpecProb	1490	5.0	50.2	1480	4.5
I-20	ISB-02+AB-TC	ISB+Yeast	2.5 Da	MSGFRaw	1272	5.7	25.2	1210	4.5
I-7	ISB-02+AB-TC	ISB+AT	2.5 Da	SpecProb	1320	4.6	36.6	1342	5.8
I-21	ISB-02+AB-TC	ISB+AT	2.5 Da	MSGFRaw	987	3.9	37.3	1064	6.1
Y-1	Y-Small+AB-TC	Yeast+AT	30 ppm	SpecProb	2574	1.0	50.1	2588	1.0
Y-7	Y-Small+AB-TC	Yeast+AT	30 ppm	MSGFRaw	1215	1.9	1.8	861	0.3

The spectral probability can be considered simply as "better normalized" score of the MS-GF score for this experiment [23]. Using the well-normalized score (i.e., the spectral probability) always produces substantially more resulting PSMs, with higher gains for larger databases. Furthermore, as in the search Y-7, the TDA-determined empirical FDR tended to be more accurate when well-normalized score was used.
¹Spectral probability was used to compute the FDR; ²MS-GF score was used to compute the FDR.

ISSN: 1471-2105