User talk:Gregor

Hello Gregor. Some of your edits seem beneficial, but some of them, such as http://snpedia.com/index.php?title=Rs7529229&curid=2197&diff=984391&oldid=955830 are harmful. Can you explain this? --- cariaso 05:46, 14 April 2015 (UTC)

Hi - Can you explain in what way it might be harmful? That's certainly not my intention. I am new to this, so perhaps I did something wrong?

you've duplicated (A;G) and removed (G;G). You've removed a dbSNP identifier. Both of those are set the way they are for a reason. I can't easily explain the 'why', I'd have to dig through tons of code to find all of the side effects. But those are not the correct values. --- cariaso 12:02, 14 April 2015 (UTC)

It was not intentional - I'm not sure how that happened. It looks right to me (I see an AA, AG, and GG). I'm not sure exactly what's wrong, or how to fix it.

I see many good edits from you, and appreciate them. I do not want to discourage you, and instead welcome you to continue. I think I see part of the problem. This edit

http://snpedia.com/index.php?title=Rs7529229(T;T)&diff=prev&oldid=984345 in particular shows that 'you' removed 3 rsnums. And I'm confident that wasn't intentional. Instead you were probably being hit by the SNPTips Bug. If so, the only advice I can offer is to disable SNPTips as they've been aware of the bug for years and have not released an updated version. As for "It looks right to me (I see an AA, AG, and GG)" look the the 2 columns of before and after at this edit http://snpedia.com/index.php?title=Rs7529229&curid=2197&diff=984391&oldid=955830 You should be able to see that geno3 was edited. I've since fixed both of those damaging edits, so you do not need to do anything to fix them. Please feel free to continue improving SNPedia, and I will continue to review your edits. --- cariaso 18:50, 14 April 2015 (UTC)

Thank you for fixing that for me - and yes, I am in fact using snptips in firefox. I was not aware there was a bug. If I recall correctly, there was a problem where one of the 3 genotypes was on the wrong strand (AA,GG, and TT, with no AG) I had some difficulty editing the TT to make it an AG, so that is almost definitely where the problems came in. I made sure it looked right to me when I was done - in the future I'll make sure it looks right in another browser or w/o snptips.

I'm rather new to the wiki editing - my experience is more in the science than in the tech. I'm figuring this out as I go along.

we all are --- cariaso 19:53, 14 April 2015 (UTC)

Good catch. confirmed. I've updated the site to reflect your discovery. --- cariaso 02:14, 14 June 2015 (UTC)

Thanks. I didn't know about opensnp - that's a very useful resource. I couldn't find any frequency data in either dbsnp or the 1000 genome project, but I started to notice that this seemed to be called way too often, there were virtually no homozygotes of either type (impossible for any MAF), and it was associated with Mendel errors in families. Perhaps the magnitude should be lower? While the mutation, if actually present, would be a highly significant finding, it's questionable whether or not a SNP called with such low specificity has any business being assigned such a high magnitude, especially in heterozygotes. I'm not sure what the snpedia convention is for handling this type of situation - i.e. a severe mutation but a very low specificity test. At least your description update should prevent a lot of undue concern.

This is probably due to variable copy numbers for this gene and its neighbors (complement 4, tenascin x) in this region of chromosome 6 and the nearby presence of a pseudogene (CYP21A1P). Due to sequence homology, there is often misalignment during meiosis, creating insertions and deletions of entire genes. One paper found a lot of false positives testing for Q318X when it was associated with HLA-Cw6, although the numbers were nowhere near this high (under 15% for carriers). I'll try to find some data on whether the nearby CYP21A1P pseudogene often carries this null mutation, which would make it very difficult to detect with SNP typing. It might be possible eventually to predict which tests are correct through LD with certain HLA's (mostly HLA-B/C) - at least if we can get good SNP markers for these, which is notoriously difficult to do.