12.07.2015 Views

View - ResearchGate

View - ResearchGate

View - ResearchGate

SHOW MORE
SHOW LESS
  • No tags were found...

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

172 Ho et al.Fig. 1. Artificial example of joint differential expression (shift) of a pair of genes. Theaxes are the measured expression levels of the two genes. Grey and black circles representsamples from two phenotypes (say normal and cancer tissue). The inner panelreflects the joint distribution; the outer margins display the univariate marginal distributions.The dashed lines represent the first principal components, conditional on the phenotype.For this data S cross= 0.01 and S shift= –0.29.phenotype: if the sum of their expression levels exceeds three, mostly thegrey () phenotype is observed. However, neither of the two genes aloneshows a strong association with the phenotype. The sides (margins) of the figureshow the 1D gene-specific distributions (technically referred to as marginaldistributions). These would have been used in a one-gene-at-a-time testingapproach, and thus both genes would have been unlikely to be selected. Thispattern is generated by the combination of a relatively high correlationbetween the genes, and a shift in the sum of the expression levels (in this caseowing mostly to gene 2) across phenotypes. Therefore, this will be referred toas a “shift” pattern. A biological mechanism leading to this pattern may occurwhen two genes are substitutes in a molecular process that is closely linked tothe phenotype. A complementary case occurs when two genes cluster aroundtwo positively sloped lines, with a shift in their difference across phenotypes.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!