Athurva Gore/LabNotes/ExomeReseq/2009-5-13: Difference between revisions
Jump to navigation
Jump to search
>Ajgore (New page: =iPS Cancer?= * Obtained Illumina Array data from Dr. Zhang * Used SAM and DAVID to try and identify GO Terms in differentially expressed genes. * Say we want a desired FDR of 15% or lower...) |
>Ajgore |
||
Line 26: | Line 26: | ||
* Should help find any genes that do not match normal Fibroblasts or hESC in iPS cells. | * Should help find any genes that do not match normal Fibroblasts or hESC in iPS cells. | ||
* Should help find some new behavior. | * Should help find some new behavior. | ||
==Procedure== | |||
* First, run SAM on Illumina sequencing data for just iPS and hESC. | |||
* Extract differentially expressed genes such that the FDR is 0.05 | |||
** Turned out to be a delta of 1.7 with 4365 differentially expressed genes. | |||
* Extracted data for just this gene subset from Illumina data, and only for FIBROBLASTS and IPS. | |||
** Saved file in NewMethod folder as csv | |||
* Imported into R, ran SAM again. |
Revision as of 22:41, 13 May 2009
iPS Cancer?
- Obtained Illumina Array data from Dr. Zhang
- Used SAM and DAVID to try and identify GO Terms in differentially expressed genes.
- Say we want a desired FDR of 15% or lower; since we are just doing preliminary.
- SAM's multiclass feature is not useful here; it will call everything that is fibroblast-only "significant," leading to numbers that are far too large.
Very Lenient
- Very lenient analysis results in a large amount of GO Terms and GO Clusters.
- Delta of 2.3 was used for IPS_HESC
- Delta of 0.1 (very low, since figured a lot of genes would need to be called as "not similar")
- Will upload DAVID results to Wiki.
- Lots of very interesting GO terms and GO clusters
- However, inspection of several reveals that these values may be too lenient. Lots of things are called as significant that are not very.
Very Stringent
- Limited ESC_IPS differences to 475, limited GO differences to 250.
- In this case, only 3 or 4 clusters were obtained at all (only 23 diff. expressed genes)
- Most of them have very low scores...probably too stringent here
New Method
- It seems SAM has issues finding differences between fibroblasts and IPS that are meaningful
- Because so many genes are differentially expressed already, very hard to choose a proper delta value.
- FDR is predicted to be very high.
- Instead, will find differentially expressed genes between IPS and HESC first.
- Only look at these in Fibroblasts, then run SAM!
- Should help find any genes that do not match normal Fibroblasts or hESC in iPS cells.
- Should help find some new behavior.
Procedure
- First, run SAM on Illumina sequencing data for just iPS and hESC.
- Extract differentially expressed genes such that the FDR is 0.05
- Turned out to be a delta of 1.7 with 4365 differentially expressed genes.
- Extracted data for just this gene subset from Illumina data, and only for FIBROBLASTS and IPS.
- Saved file in NewMethod folder as csv
- Imported into R, ran SAM again.