Table 3. Confidence in Algorithm Detected Peptides Depends on Parent Protein Expression and Distance from Decoy FDR Threshold.

		SEQUEST			OMSSA
		conforming peptidesb	nonconforming peptides	conformance scored	conforming peptides	nonconforming peptides	conformance scored
(A)	log(PE)a
	undetected	669	115	85.3	558	74	88.3
	x ≤ 3	171	117	59.4	102	62	62.2
	3 < x ≤ 4	668	228	74.6	652	169	79.4
	4< x ≤ 5	1467	258	85.0	1446	215	87.1
	x >5	1093	166	86.8	850	133	86.5
(B)	scoring confidencec (d)
	d ≤ 1	724	420	63.3	431	270	61.5
	1 < d ≤ 3	1291	241	84.3	741	135	84.6
	3 < d ≤ 5	938	135	87.4	787	106	88.1
	5 < d ≤ 7	597	66	90.0	642	93	87.3
	d >7	561	43	92.9	1123	101	91.7

^aProtein expression (PE; estimated protein molecules per cell) based on large scale Western analysis.¹³

^bPeptides are counted once per LC–MS/MS run even if detected by both the b/y and a/b/y ion screens and, in the case of OMSSA, even if detected by multiple parameter sets.

^cScoring confidence d = −log₁₀(PSM_score/FDR_threshold). For OMSSA, the algorithm PSM score used was the e-value. For SEQUEST, implemented in Proteome Discoverer 1.2, we used the probability outputs to compute a PSM score = 10^{(probability/–10)}.

^dConformance scores are significantly depressed for lower abundance proteins (chi-squared: SEQUEST, p < 6.18 × 10^–36; OMSSA, p < 3.46 × 10^–20) and for lower d scores (SEQUEST, p < 7.23 × 10^–80; OMSSA, p < 6.25 × 10^–72)