# Investigator: Jim Goldenring

## Description: Cell lines with control and treatment with DMP-777 on day 1 and day 3 (independent samples on two days). Sample size is 4 controls, 4 day 1 and 4 day 3 (need to check on this, at one time there were only 3 samples run on day 4 but one more sample was added). Reference paper American Journal of Physiology.

### Tasks

#### Normalization (Zhouwen).

Used RMA (in Biocondutor package) to normalize these data. Completed 1/25/06.

#### SAFE analysis (Aixiang).

Need to save the local class results so that we can look at specific differences (e.g., we need to examine the specific pair-wise comparisons). Question: Can SAFE do this? Does it perform a permutation ANOVA? If not, we will need to look at the following 2-sample tests: control versus day 1 and control versus day 3.

#### Additional analyses (Aixiang/Bonnie).

- Need to examine ANOVA results to compare with data analysis using data that were normalized in microarray core lab - missing gene from default normalization, is it present using RMA? (TTF2/spasmolytic polypeptide - SPEM - affy probe name(s): 1428552_at, 1447119_at, 1452818_at)

- Particular differences those probes for which the following pair-wise differences are seen:
- day 1 > control AND day 3 > 0
- day 1 > control NOT day 3 > 0
- day 3 > control NOT day 1 > 0

After 2/1/06 meeting it was decided to approach this analysis in two phases. The first will be the examination of probe level (local) analyses:

- ANOVA for overall difference, if p < 0.05 then proceed
- T-test for paired differences between conrol and day 1 and control and day 2, if BOTH have p < 0.05 then output and annotate

this will be compared to analysis based on SAFE analysis (both local and global) -- here we may need to make sure we have the directional information for each pairwise difference, so we can compare.

Prior analysis done by microarray core (normalized using MAS?) with the criteria of day 1 > control AND day 3 > 0 generated a final list of 466 probes (from 7,187). When the SAFE list is generated we will rank the products of p-values for the pairwise differences (multiply the p-values for each comparison) and plot the ranks (analysis 1 versus SAFE) to see what type of correspondence there is between these analyses. Additionally, we will generate a Venn diagram of the overlapping genes.