observed weighted vairant count

[pnrobinson]

For some genes, we find multiple rare predicted pathogenic variants in the population background. For instance, KCNJ18 was found to have ten variants and got an Observed weighted deleterious variant count: 21.48. λdisease=1. λbackground=0.1000.
This is clearly an error and we should limit the number of variants that go into this score.
The conservative way of doing this would be to take the max for AD diseases and to take the best and second best allele for AR etc. In all likelihood, if we find this many alleles, there gene is simply variable and it would be better to downweight, but this assumption might not always be true, so I would prefer the first option.