SAGE significance

Rork D. Kuick rork at umich.edu
Mon Jun 16 08:25:23 EST 1997


In "Gene expression profiles in Normal and Cancer Cells" (Science, 23 may 97,
1268-1272) Zhang, Zhou, Velculescu, Kern, Hruban, Hamilton, Vogelstein,
and Kinzler, sample a great many transcripts using SAGE from different
sources and then compare the frequency of transcripts.
 I cannot understand their statistical comparison - their note #8 was
difficult.  I'd like to think that it not just my problem - I perform
Monte-carlos to compute ssignificances as well from time to time.
Does "differences in expression" mean "in the sampled SAGE
transcripts" or "in mRNA molecules from which we are sampling"?  
 I do not know why they do not use conventional methods assuming
poisson or binomial distributions.  Clearly there is a "problem" of performing
thousands of statistical tests, but this this problem is not rare.
 I cannot compare how their method judges the significance of differences
compared to simpler statistics, since no where do they give an example
with both the frequencies and their "significance" at the same time.
  
1)  If you understand what they are doing, tell us.
2)  How would you do it?
    Thanks,  rork kuick



More information about the Methods mailing list