Thanks for adjusting the language to be little less alarmist, though I'll note there are actually multiple studies linking the *protein* with ciliopathies and inherited retinal degeneration in particular. What I tried to but may have failed to carry across is that the cilipathy associations are for the *protein* and not this SNP in particular. Only reason to think this SNP might be related is because Swift/PolyPhen has it as "possibly damaging" or higher. One way to stress this difference would be to move the detail specific to the gene into page for the gene, but I added the gene page only later and wanted to motivate why the SNP is of interest at all.
Worth noting is that DNA microarray technology relies on calibrating genotype calls by clustering from results, which is difficult for low-frequency alleles, being one reason they usually aim for SNP's with at least 5% MAF. Rs13225917 is ceratainly rare enough that it could have high genotyping error rate, especially relative to true calls. However I do have a mother-child pair genotyped on different chips where the haplotype is preserved, so there's reason to believe it isn't a junk SNP. Not that it neccessarily means anything with regards to its expression, of course.
The "usual" caveats vary: The rarer the SNP, the higher the relative chance of false positive, even a damaging mutation may affect a vestigal function of the gene that is no longer relevant to humans, and important clinical pathways are often redundant requiring failure in multiple genes to result in phenotypical expression. This is why GWAS is still the main tool of finding associations, after all, but it would be difficult to have large enough test group to have enough statistical power for a rare variant, even if genotyping and processing powere weren't an issue so they focus on common variants. Plus the phenotypic variation, if it exists at all, could be one no-one expects or studies, of course. --Donwulff (talk) 13:30, 4 June 2014 (UTC)
- We're in complete agreement on this. It is interesting and useful, especially for researchers, to point out at the protein level (which in SNPedia really mostly means the gene level) associations with disorders, either as individual disorders or classes of disorders. For particular SNPs in a protein, though, we are taking the position that although anyone is welcome at any time to make predictions based on PolyPhen or the like, those predictions shouldn't wind up in the genotype pages for that SNP, since the average Promethease user is generally not well served (and gets spooked) by seeing an unsubstantiated prediction pop up in their personal report. [Incidentally, although it should be possible to link out to a SIFT or PolyPhen prediction quite readily from any main rs# page in SNPedia by heading to the Ensembl link found in the right sidebar box, and then looking at the table that results from clicking on the "See all predicted consequences" link, unfortunately, the SIFT and PolyPhen data is often missing.] Eventually, as PolyPhen and other programs improve AND more and more users have genomic or exomic sequence instead of chip data, it will probably make sense to have a completely new section in Promethease asking users to self-report if they wish about SNPs having (only) predicted consequences of note. Greg (talk) 17:19, 4 June 2014 (UTC)