I realize it's a big topic, but I'd be curious to hear people's opinions about the best ways to rank genetic variants obtained from an NGS procedure by their probability of being disease causing. I'm aware of SIFT, PolyPhen, MutationTaster, etc., but these all seem to give conflicting results much of the time. What tools do you use? What techniques or algorithms have you developed? Is it even possible?
It's not surprising that those algorithms give conflicting results. This is because the disease associated genetic variants are quite different in different populations. You will see different results if you use data from different populations. So better to use your sample to derive the associated genetic variants.
If you have a genome wide association data, use Chi-Square test to rank the genetic variants. If you don't, use SIFT, PolyPhen, MutationTaster, or MSRV for coding regions (http://tingchenlab.cmb.usc.edu/msrv/), or SinBad (http://tingchenlab.cmb.usc.edu/sinbad/) for intronic, promoter, and coding regions. These tools are based on evolutionary conservation to rank genetic mutations.
Hi, my response perhaps is from a different perspective. You can try G2D , http://www.ogic.ca/projects/g2d_2/. There is also a gene prioritization portal http://homes.esat.kuleuven.be/~bioiuser/gpp/ . Endeavour is a nice tool. If you can define your phenotype of a disease by a list of genes, then the list of unknown genes that you have is scored by similarity to the defining genes.
Have a look at the VAAST tool. It's a probabilistic disease gene finding tool: http://www.yandell-lab.org/software/vaast.html
Thanks for these ideas - is there any word on how accurate these things are? Maybe it's premature to start into this, but it would be nice to have some way of comparing these various tools, and saying "this one works the best", at least in certain circumstances.
The original VAAST paper (http://www.ncbi.nlm.nih.gov/pubmed/21700766) had a number of comparisons along those lines. A new paper which will be submitted in the next few weeks has a much more extensive comparison of VAAST as both a variant prioritization tool, and a disease gene prioritization tool. You could check back on the VAAST website mentioned above or follow the VAAST mailing list to get updates on publication of the new paper.
- Badges
- Science method