The utility of MAS5 expression summary and detection call algorithms

BMC Bioinformatics

Table 2 Affycomp scores for RMA, GCRMA, MAS5 and MAS5 filtered by detection call

	rma	Gcrma	mas5	filtered mas5
Median SD	0.07	0.06	0.29	0.08
null log-fc IQR	0.13	0.04	0.47	0.01
null log-fc 99.9%	0.40	0.61	4.01	1.25
Signal detect R2	0.90	0.91	0.91	0.88
Signal detect slope	0.68	1.00	0.77	0.77
low.slope	0.20	0.25	0.58	0.25
med.slope	0.71	1.13	0.73	0.79
high.slope	0.80	0.97	0.77	0.77
Obs-intended-fc slope	0.68	1.00	0.77	0.77
Obs-(low)int-fc slope	0.31	0.48	0.64	0.44
low AUC	0.43	0.35	0.00	0.19
med AUC	0.85	0.86	0.00	0.18
high AUC	0.93	0.86	0.00	0.18
weighted avg AUC	0.53	0.48	0.00	0.19
AFP, call if fc > 2	1.71	3.30	2863.36	73.73
ATP, call if fc > 2	32.91	35.40	35.45	34.18
FC = 2, AFP, call if fc > 2	0.24	1.95	2785.10	63.52
FC = 2, ATP, call if fc > 2	10.81	19.57	14.33	12.86

Metrics are described in detail in (Cope et al. 2003) and briefly in the methods. The last four rows record the number of True and False positives. Perfection for the first three rows is 0, and 1 for the next 11 rows. Filtering by detection call significantly improves the performance of MAS5 on the Latin square data.

ISSN: 1471-2105