QUESTION: Any thoughts on this piece of research, forwarded to me from a patient noting a 40% inaccuracies in SNP assessments? That would be quite high based on the changes we might recommend to patients based on the reported SNPs. “According to information derived from the GAOs report on Consumer Genetic Testing, a SNP analysis can be as high as 40% inaccurate. SNP is one piece of information that is different from what is expected to be there in respect to the two strands of DNA, whereas the full sequence identifies multiple elements and a full picture of the genetic information. As well as sequencing through the SNP instead of just trying to identify a single nucleotide change.”
ANSWER: Good question. I don’t see the source of this ‘research’ but I have seen the following:
1. A SNP may represent one rsid or as in the case of some commercial SNP tests a group – so if one looks at the group one gets an answer potentially different from the other looking at one rsid. Not two different answers but two different questions. Most consumer ones are one rsid = one SNP affairs.
2. A testing company can suck at what they do. We have seen this in lots of lab areas and then the issue isn’t the idea of the test but the company doing the testing / primer use etc. I generally (when used in the context I use them) see the SNP results from 23&me raw data run via MTHFRSUPPORT or even GG to be useful from the standpoint of clinical decision making and cause (SNP etc) matching effect (Tx) but again I don’t treat SNP’s but do use them to inform Tx/Rx.
3. If the idea of not treating the SNP, but factoring that into the bigger picture and using the multi-step dose strategy I teach in my classes (along with baseline support first etc etc) is used for therapy nobody really will get overdosed anyway, and the first rule is if they do aggravate you take it away, drop the dose or go back to baseline support etc. SO if done correctly the chance of harm is very low.
4. In the future the testing and labs need more validation for sure – but based on the way the GAO ran the assessment a lot of error can be induced and reported. So I’m sure some error exists but I’m skeptical about true error rates in the double digits (but again, some labs suck…)