Entering edit mode
6.9 years ago
Sharon
▴
610
Hello Everyone
Happy new year ! We are working on very very rare variants. We are looking for multiple or combinatorial variants or mutations that could be a possible cause for some very rare phenotypes. For example we are interested in finding variant A + variant B that cause a phenotype. A could be a very rare variant but not sufficient in itself to cause the phenotype. B could be a more common variant, but A + B cause the phenotype.
Is there any tools for finding that or I should write my own code for that? I usually use Annovar or VEP for annotations
Thanks
I recently analysed a complex and very rare disease about which there is minimal information and on which whose disease mechanisms clinicians the World over disagree. Faced with the issue that there was no clear mutation causing the disease (based on information from hundreds of samples), I developed my own risk score algorithm based on a Bayesian logistic regression model and using various other information in order to calculate the scores. The resulting model actually has very high predictive power for disease samples. I can't go into any more specifics as the work is protected.
Neither Annovar nor VEP deal with these issues. Almost all tools in current use have been designed with GWAS in mind, i.e., SNPs of a reasonably high MAF in mind.
A good starting point for you, though, may be looking at Burden testing.
I did all types of rare variant association analysis, burden test, adaptive burden test, SKAT, SKATO, ..etc. Nothing interesting so far found.
Okay, that makes sense. Start thinking about region conservation (in which the variant is found), predictions from multiple in silico tools, and the comparable frequency of the variants in cases versus controls.
This is an active/future area of research and open to new methods. Most diseases, we now understand, are based on complex genetics. You still find many people who expect just a single gene to explain everything, though.
Hopefully others here can contribute.
Ah, I appreciate if you can explain more the region conservation and comparable frequency. Like I compare the variant in case and control, keeping in mind possibly carrier, penetrance, ..etc. What else do you mean? Thanks