Stat Bayesian Analysis Notes

  1. For each of the 8 animals M=log2(D/U) was computed for each of the 27,648 spots and normalized using global lowess normalization (from the R sma package). Here D=Disturbed flow region and U=Undisturbed flow region.
  2. The two M's for duplicate spots on each arrays were averaged, resulting in 13,824 average (normalized) M v alues.
  3. The values from (2) were fed into the stat.bayesian routine of the R sma package to generate rankedBylods2.txt. Recall that this method simply outputs a ranked list of all genes according to decreasing lod score, where the lod score is the logarithm of (probability of being differentially expressed)/(probability of not being differentially expressed) given the data. Whether or not this is up or down regulation can be seen by looking at the sign of the average value of M for that spot (positive being up-regulation). This is just a ranked list, it is up to the investigator to decide how many of the top genes (s)he wants to follow up on. Note that at the end of the list are those genes for which lods could not be computed (NA) for a variety of reasons.