Brief Description
The context here is that of gene mapping. Basically, there is a phenotype, such as height, heart attack, alzheimer’s, etc- can be continuous (height), or binary (has had heart attack Y/N). And there are two relevant questions from genetics:
- What percentage of this phenotype is explained by genetics (for height, it depends more on diet, so genetics explain a small portion of the variation in height, and environment (we use environment with the definition from genetics here, meaning “anything not genetic”)- diet here, explains a lot of the variation in height.
- Which genes are responsible for the genetic part? So if we determine that say, Huntington’s disease or hemophilia has a genetic component, then which gene or genes are responsible explaining the phenotype?
This project has to do with (2) above. Further, all it does it change the class of algorithms that we use. The original description is here.
Specific Objective
Write a WGS algorithm which is better than the current algorithm. The genetic model doesn’t change, only the underlying algorithm from numerical linear algebra which we use to solve the model changes.
Product/Results
At the end of the day, I would like to have a C routine (based on LAPACK) and/or an R library which implements the algorithm and is available for download. I would also like to get a publication of some sort out of it (I do need a real job to give me time for this stuff).
Status
Running: In progress, trying to give it a couple hours a day. With the move just about over, and no part time work on the docket, should be going on over the next couple of weeks.
Current Research
I will put a draft here when I have something submittable- I don’t want my ideas to get poached too easily, but I will disclose all, have no fear.
