However, the proposed gammaMAXT algorithm offers a general significance assessment and multiple testing approach, applicable to any context that requires performing hundreds of thousands of tests.
The aim of the adjustment is to control the type I error rate defined with respect to the multiple testing approach.
The identification of such individuals has failed in several provocation studies so far (reviewed by Rubin et al. 2005) and would require a multiple testing approach to reduce the likelihood of a correct rating by chance.
A multiple testing approach was applied, with the criteria for inclusion being set to a P-value of <0.05 for the t-tests and a q-value of ≤10% in the SAM analysis, when comparing trisomy 8 with disomy 8 (CT8M only) and when comparing trisomy 8 with disomy 8 (CT8M and references grouped together).
A multiple testing approach identified 25 protein-coding genes, one small nucleolar RNA, and one miRNA that were significantly overexpressed and four genes and one miRNA that were significantly underexpressed in the trisomy 8 cultures as compared with both the disomy 8 and reference cultures.
For further discussion on multiple testing approaches, see Dudoit et al. (2003).
Nevertheless, the possible merits of test combinations in novel multiple test approaches including additional serological markers, might be a promising area for further research.
Thus, under the AU test, the complete LGT footprint is 286 ORB+ (as above) plus 329 ORB− for a total of 615 gene sets (42.1% of 1462), 29% greater than the 476 identified using the multiple-test approach above.
The 489 discordant gene sets include all 342 discordant sets identified using our multiple-test approach, demonstrating that our estimate of the extent of whole-gene transfer in prokaryotes is of high confidence and perhaps conservative.
Even though we had corrected for multiple testing, this approach may have been ineffective to correct for false positives.
Furthermore, our simulations suggest that, even without correction for multiple testing, this approach will lack power compared to sex-differentiated tests of association when causal variants have allelic effects in both males and females, but which differ in direction and/or magnitude between them.
