After some months of work, we had our article accepted in bioinformatics and the accompanying  software GLASCOW released (available on the GIGA website).

Basically, the software perform a generalized linear mixed model analysis on (preferably) hidden states (obtained e.g. with PHASEBOOK) after a correction for population stratification. This approach has already been successfully applied in : Durkin et al, Sartelet et al, Dupuis et al.

I contribute to the software with some optimizations and code rewriting. The code is thus a mix of several coding practice and style, I must admit that to some extent this can made the code hard to follow.

Seen on this blog

Concerning programming, several details of the package are available on the blog  :

  1. The general framework of the package was set by the script
  2. I use command argument as described here 
  3. The use of Lapack is conditional, and therefore I add some preprocessor instructions in the code
  4. In order to get info on date of compilation I also add some preprocessor instructions

…coming soon on this blog ?

Others key features are :

  • A parallelized reading of the files (we rely on the use of one phase/hidden states file per chromosome, obtained with PHASEBOOK), this trick plus a neat storage of the partial relationship between individuals allow to obtain great speed improvements and lower memory consumption.
  • The computation of score on residuals allow both a general speed-up and an ability to restart analysis only from these last steps.
  • Permutations are done in parallel, nevertheless this last part was very tricky to implement, and I must says that the speed-up didn’t fulfill my expectations….I’ll try to see how to fix this.
  • We add an option for outputting results in a “.assoc” file for a better viewing experience in IGV.

So even if the code still contain some old fashion Fortran (goto etc…) the adventure was fun, inspired me some posts, give me some experience in glmm and add one article to my publications list ! Not so bad, isn’t it ?

  1. No comments yet.
  1. No trackbacks yet.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: