I looked at APOA1 and found rs670 with MAF=18%, minor allele = T
This gene was found to have in a sample of 62 people :
No mutation = 53.45%
Homozygous = 3.45%
Heterozygous = 41.38%
a) Is the observed difference between MAF and Heterozygous considered significant and warrant this particular gene for further investigation? If you also can point me to relevant information on how you make these calculations it would be great.
There's a couple simple formulas for calculating expected frequency (MAF = Minor Allele Frequency):
minor homozygous = MAF x MAF
heterozygous = (1-MAF) x MAF x 2
major homozygous = (1-MAF) x (1-MAF)
If your sample is primarily Caucasian (it probably is, due to the availability and use of the 23andMe test), then it's most meaningful to compare it to allele data from a European sample. Otherwise it might look like there is significant variation, but the variation between different ethnic groups may simply be down to a matter of chance and historical separation.
dbSNP is a good source of both very large overall samples, and various ethnic samples. If there isn't a large sample (100+) with similar ethnicity to your sample, then the sample might not be a very good one to draw any conclusions from. EUR samples are likely to be a lot better than CEU samples, due to the CEU samples being from Caucasians in a single state in the US where a smaller religion is very dominant, and has resulted in some insularity that is unlikely to represent wider populations.
If we scroll to the bottom of
https://www.ncbi.nlm.nih.gov/projects/SNP/snp_ref.cgi?rs=rs670 we can see that the 2nd entry under "Population Diversity" is for EUR with a size of 1006. 1006 is the number of alleles, not people, so it's only 503 people. Which is still pretty big, for a specific ethnic sample. To the right of that entry, the percentage of alleles are listed, with 15.4% of alleles being the minor allele.
There's also a smaller European sample of 48, with a drastically different rate of 31.3%, which is a good illustration of the problem of working off of smaller samples
As well as the amount of variation that can be seen in a completely normal sample, down to random chance when a great many samples are taken. Basically, it's illustrating why even a big difference in your sample compared to another sample might not mean anything. When hundreds or thousands of variables (SNPs) are compared, it's guaranteed that some rates will randomly be different.
With the equations listed above, we can determine the expected rate in your sample:
Minor homozygous = 0.154 x 0.154 = 2.4%
Heterozygous = (1 - 0.154) x 0.154 x 2 = 26.1%
Major homozygous = (1 - 0.154) x (1 - 0.154) = 71.6%
So that's very similar for the minor homozygous, but fairly different for heterozygous. But if it was particularly relevant, I'd expect to see both a much lower MAF, and more of an increase in being homozygous for the minor allele. Common heterozygous mutations are rarely having a sizable impact.
b) Do you agree that combination of Gene SNPs may or may not result to a certain phenotype?
Yes, to a limited extent. Compound heterozygous missense mutations (which can't be determined with 23andMe data) certainly can. But I'm not convinced regarding haplotypes. Those studies typically find no direct association between increased prevalence of a SNP and a condition, then start combining SNPs until they get a different prelevance compared to controls. I haven't seen any studies making proper statistical corrections to account for the extreme likelihood of randomly generating a meaningless correlation.
But the broader problem is that genes generally don't really seem to work in a way which supports a model where several SNPs have literally no impact on their own, but do have an impact when they're all present together. If a study is properly powered, it should be able to find even those individual tiny impacts.
And haplotypes involving different genes seems like an especially absurd proposition. If the SNP on each individual gene has absolutely no impact on its gene, then they are making a completely normal protein product. There's no conceivable way that those normal proteins will somehow interact differently. Actual differences can certainly aggregate, but there has to be some difference to start with.
I'm also not aware of any studies which have demonstrated a functional difference in the protein created by a haplotype composed of SNPs which have no impact at all on their own. Instead, haplotype studies are comparing the prevalence of SNPs in people with a certain physical, personality, or behavioral trait, against normal controls. Good genetic research should go deeper, and demonstrate how the genetic variation results in the difference in phenotype. Most researchers playing with SNPs lack the skills to do that. But at the very least, there needs to be independent replication with new samples from different patients and different controls - this will usually make false-positives evaporate.
c) In the latest CFS/ME symposium G-Protein coupled receptors were mentioned. Do you know if any relevant SNPs were given?
It's "ME/CFS". But no, I didn't watch the OMF conference, since I have trouble processing informational videos, and I do much better with text. Additionally, claims at conferences lack subtlety and context that is included in the published version, so I tend not to get too excited about that sort of thing anyhow. Better to wait for the paper.