False discovery rates in spectral identification

BMC Bioinformatics

Table 12 Comparison between the single-pass search (the search Y-1) and various two-pass search methods (the searches Y-10 to Y-13)

Search#	Spectra	Database	PMTol			MSR²	EmpiricalFDR fixed			FactFDR fixed
				IsTwoPass	2th decoy¹		N _target	FactFDR(%)	p-value(%)	N _target	EmpiricalFDR(%)
Y-1	Y-Small+AB-TC	Yeast+AT	30 ppm	No	Rev	N/A	2574/1988	1.0/1.3	50.1/22.7	2588/1759	1.0/0.5
Y-10	Y-Small+AB-TC	Yeast+AT	30 ppm	Yes	Trad³	No	5361/5744	15.9/20.1	0.0/0.0	3260/2655	0.6/0.3
Y-11	Y-Small+AB-TC	Yeast+AT	30 ppm	Yes	Trad	Yes	4114/3925	7.3/10.1	0.0/0.0	3102/2320	0.9/0.5
Y-12	Y-Small+AB-TC	Yeast+AT	30 ppm	Yes	BK⁴	No	3529/3089	1.2/1.0	24.9/45.0	3262/3074	0.7/0.9
Y-13	Y-Small+AB-TC	Yeast+AT	30 ppm	Yes	BK	Yes	3137/2514	1.1/1.1	40.1/39.3	3103/2521	1.0/0.8

For the searches Y-10 and Y-11, the traditional second pass decoy database was used to estimate FDR (see text). For the searches Y-12 and Y-13, the decoy database proposed by Bern et al. [25] was used. Also, for the searches Y-11 and Y-13, the matched spectrum removal (MSR) step was used.
Low Fisher p-values in Y-10 and Y-11 illustrate that using the traditional second pass decoy database results in significant underestimation of the true FDR.
¹The decoy database for the second pass search; ²Whether the matched spectrum removal step was used; ³The traditional decoy database; ⁴The BK decoy database.

ISSN: 1471-2105