In the first video about sensitivity analysis for Mendelian Randomization, it mentions that people often exclude palindromic SNPs because of the possibility of strand issues (but that sometimes things like allele frequencies can give you a clue whether that’s a concern or not).
I wanted to ask - are strand errors still a concern with modern genotyping technologies?
As far as I understand, on a genotyping chip, a SNP is found for genotyping based on the sequence of variants flanking it. So one probe might be ACGAGTTATAG and the other might be ACGAGATATAG
And if you’ve got sequencing data, it’s assembled by aligning reads to a reference, which also involve blocks of nucleotides and not just single SNPs.
In both cases, it seems like a strand error would not be possible unless:
- The sequence flanking the SNP was also palindromic, eg. ACGAGACTCGT (which, in reverse-complement, becomes ACGAGTCTCGT) - the probability of that would get lower the longer the sequences surrounding the SNP were.
- In the case of a genotyping chip, the chip company made a mistake when designing probes and made one that matched the wrong strand.
- Some other explanation I haven’t thought of.
Are strand errors actually still common enough with modern genotyping technologies that we have to worry about them? Or are palindromic SNPs mostly not a concern any more?