Robust differential expression analysis by learning discriminant boundary in multi-dimensional space of statistical attributes

Table 4 Compares the average numbers of true DEFs identified in different distribution categories

Detected by	Total	NB	Gaussian	GMM (2 or 3 components)
DC^T	252.79 (3.18e-35)	115.60 (3.16e-27)	55.83 (8.12e-10)	81.36 (2.24e-21)
DC^NB	228.55 (8.96e-56)	104.16 (7.39e-55)	48.33 (7.87e-39)	76.06 (7.01e-35)
DC^T+R+NB	266.22	121.88	57.54	86.80

Compares the average numbers of true DEFs detected by DC^T, DC^NB, and DC^T+R+NB in different distribution categories under the simulation test setting: 6 vs. 6, G ⁵⁰⁰₅₀₀ , and the target FDR < 0.05. DC^R is not displayed because it detected no DEFs. The numbers in the parentheses are the paired t-test p-values indicating how significant DC^T+R+NB outperformed the corresponding single-attribute DC configurations.

ISSN: 1471-2105