Selecting minimal gene set (with co-mutations) for designing a panel

Question

[SLP]

Hi everyone, I've a list of SNV's (in about 70 genes) in a cohort ~200 patients. Many of these SNVs are co-mutations. Now, I wish design a gene panel that would cover >80% of patients. One strategy I can think of is to order the SNVs in decreasing order of their frequency (in patients) and start excluding the least frequent ones from the bottom 1 at a time to see what's the minimum # required to cover 80% patients.

But this strategy doesn't account for co-mutation. e.g. what if 2 SNVs have high frequency but always occur together. We can clearly just one of these on the panel. But by implementing the above strategy I'll end up keeping both of these.

I need suggestions to tackle this problem more efficiently.

Thanks SLP

[Kevin Blighe]

Hello, you could instead consider a different approach and aim to define 'haplo blocks' or 'haplotypes', i.e., groups of variants that are in high linkage disequilibrium (LD) with each other. If you set the LD threshold high enough, it would mean that, statistically, you could infer the genotypes of all variants in a given haploblock even if you knew the genotype of just 1 variant in the same haploblock.

The only issue with this idea is that it was designed or genome-wide data. How sparse are your variants across the genome?

Kevin