Justification For Use Of Hapmap As Reference For Snp Analysis

Question

[user1202664]

For SNP microarray experiments, why do Affymetrix justify the use of pooled HapMap data in order to make copy number and loss of heterozygosity inferences on experimental data that is submitted to them? Ideally, you would submit patient-matched normal/reference samples (for solid tumors for example, this would be blood) to them. However, the alternative to this is that they use 250 samples from HapMap as reference. What is the justification for this?

Answer 1

[Jorge Amigo]

I'm pretty sure this is simply because the HapMap samples have been through such a deep clinical and phenotypical investigation that all the metrics that they've already calculated (heterozygosity patterns, LD blocks, population frequency and so on) are very easy to share, and for that reason you can use these samples as your control references when you deal with tumor cases in your experiment and you want to calculate comparison metrics such as loss of heterozygosity. sure that comparing normal vs tumor dna from the same sample would be the best thing to do (affy does not stop you from doing that, they simply offer you the option of not doing it), but then you would have to double genotype, and therefore perform all the quality control and downstream checks twice. this way (comparing with HapMap) you only have to type your case samples, which is of course cheaper and faster, plus the extensive work on the HapMap project side gives you a good standard for your quality control measurement.

[user1202664]

I understand the cost and time benefits. I was speaking from the purely biological perspective. Does using HapMap invalidate your ability to say whether a particular locus has a loss of heterozygosity? Or are inner workings of the calling algorithms (PennCNV, ASCAT, etc.) on the HapMap controls sufficient to do the biology justice?

[Jorge Amigo]

unfortunately I can't go that further. in my opinion HapMap data is "just" a useful resource, but if you have cases and controls and you have the time and the money to test both hypothesis, surely it'll be more accurate.