Department of Biology

Renn Lab

Fish 'N Chips: Comparative Genomics and the Evolution of Behavior in African Cichlid Fishes.

PAGE

SOFTWARE - R scripts

Optimal data normalization:

After repeatedly asking statisticians how to normalize microarray data and being frutstrated by the response of "it depends on the data", we partnered with Albyn Jones (Reed Mathematics) to develop an R-script that would run multiple combinations of background correction and normalization and compare analyzed results. We take advantage of replicated features on the microarray to identify the optimal normalization strategy as that which results in the least variation between replicate array features yet maintains the most variation between features representing different genes.
These scripts are available to download. (note this is specific to our cichlid array, please contact us for assistance tailoring this script to another micorarray)
The poster that summarizes the work of Gavin Brown is available to view.

 

 

Test of Equivalence for Microarray Data:

After repeatedly being annoyed at publications that claim specificity of gene regulation under one condition simply because the same gene was not significantly regulated in the other comparison, we worked with Albyn Jones (Reed Mathematics) to develop a script in R that would work with LIMMA output files in order to actually test for "equivalence" of gene expression using the Two One Sided Tests method (TOST). The difficulty still lies in determining epsilon, the effect size determined to be equivalent. Nonetheless this approach will allow one to avoid the Freshman folly of assuming that evidence of absence is absence of evidence.

There are two scripts available one to calculated pooled standard deviations used to determine an epsilon, another to do the TOST.
The poster that summarizes the work of Andrew Winterman is available to view.
Also, a rough draft of a manuscript that was written before we were scooped. :)

 

Mixed Model Analysis in LIMMA:

UNDER CONSTRUCTION

 

These R-scripts are the product of NSF funded summer REU fellowships associated with NSF-IOS Award--0818957 to Renn and Jones.