In this TP, you will compare several multiple
hypothesis testing p-value adjustments.
Here, we use the ones implemented in
We will again use the e. Coli data.
You should read in the data, compute RMA,
fit a linear model and carry out empirical Bayes
shrinkage –(see the practical from
last week.
You might just want to check with
The p-values for the mod t statistics are in
It will also be convenient to save the sort order of the raw p-values, for example:
Use
Now you can make a plot with all the adjusted p-values together.
Here is an example how you might get the raw
and Bonferroni adjusted p-values; you can add the other adjustments with more
calls to the function
Which genes have adjusted p-values less than 0.05?
(You can use
One thing that I usually do when I am reporting on data I have
analyzed is to present a 'volcano' plot (B vs. M or abs(mod-t) vs. M).
For the simple design in this data set, the M (log2 fold change)
values are in
For practice (remember that exam??), you can send me a short report on identifying differentially expressed genes in the e. Coli data. You should include a brief purpose of the investigation (i.e. to look for genes differentially expressed between lrp+ and lrp-), the linear model you are fitting to identify DE genes (including the design matrix), the p-value adjustment you are using and a short explanation, a volcano plot and a list of genes that appear to be 'interesting for followup', along with a short justification for calling them 'interesting' (e.g. FDR < 0.05). State how many genes are identified as DE using your criterion. In addition, include on a separate page a table of DE genes (or just the top 50 if you have more than 50). This report should not be more than 3 pages.
For the real exam, you will also need to include an exploratory/quality analysis and explanation of the model, fitting procedure, and quality measures as well as an explanation of RMA. If you want, you could also include these in your report, in which case your entire report will be at most 6 pages.
As usual, your report can be in English or French.
Please send your report as a pdf file,
and follow the naming convention: lastname.pdf
(e.g. my report would be goldstein.pdf).
If you email your report to me