Table 2. PPVs and NPVs for Different Combinations of Parameters and Methods (Imputation and Normalization) on Protein Level^a.

Tool (imputation-normalization)	UPS spiked data set		Tool (imputation-normalization)	Large-scale data set (4:1 fold)		Large-scale data set (10:1 fold)
	PPV	NPV		PPV	NPV	PPV	NPV
Proteus (MVL-NN)	0.97 (31 TP, total 2231)	0.998 (3 FN)	Proteus (MVL-NN)	0.62 (431 TP, total 3694)	0.972 (85 FN)	0.54 (475 TP, total 3403)	0.981 (48 FN)
prolfqua (GMI-RS)	0.91 (39 TP, total 2143)	0.999 (1 FN)	prolfqua (GMI-RS)	0.65 (544 TP, total 3043)	0.879 (267 FN)	0.54 (703 TP, total 3043)	0.938 (108 FN)
ProVision (ND-NN)	0.93 (38 TP, total 1987)	1.0 (1 FN)	ProVision (ND-NN)	0.62 (239 TP, total 2310)	0.851 (287 FN)	0.53 (400 TP, total 2310)	0.919 (126 FN)
LFQ-Analyst (QRILC-NN)	0.93 (38 TP, total 1988)	1.0 (0 FN)	LFQ-Analyst (QRILC-NN)	0.70 (263 TP, total 2108)	0.836 (284 FN)	0.59 (454 TP, total 2108)	0.931 (93 FN)
Eatomics (ND-limma VSN)	0.93 (38 TP, total 1826)	1.0 (0 FN)	Eatomics (ND-limma VSN)	0.63 (43 TP, total 1122)	0.992 (8 FN)	0.33 (49 TP, total 1041)	0.998 (2 FN)
ProStaR (DQ-GQA)	0.84 (37 TP, total 2238)	1.0 (0 FN)	ProStaR (DQ-GQA)	0.55 (304 TP, total 5521)	0.739 (1295 FN)	0.57 (504 TP, total 5521)	0.764 (1095 FN)
Perseus (GD-NN)	0.95 (38 TP, total 1946)	1.0 (0 FN)	Perseus (zero-NN)	0.65 (159 TP, total 1593)	0.950 (67 FN)	0.63 (314 TP, total 1629)	0.955 (51 FN)

^aFor all these tests, the input protein expression tables were generated by MaxQuant to perform the differential expression analysis. Only the best combination for each tool is presented. Supplementary Table 4 contains all combinations’ results. The definition of the best combination is that the higher the PPV with the same amount of TP proteins. If there is a large difference in the amount of TP proteins, the greater the amount of TP proteins, the better the combination.