Contribution of rare variants to heritability of a disease is much greater than conventionally estimated: modification of allele distribution model.

"Missing heritability" is a current problem in human genetics. I previously reported a method to estimate heritability of a polymorphism (h) for a common disease without calculating the genetic variance under dominant and the recessive models. Here, I extend the method to the co-dominant model and carry out trial calculations of h. I also calculate h applying the allele distribution model originally reported by Pawitan et al. for comparison as a conventional method. But unexpectedly, h calculated for rare variants with high odds ratios was much higher than the calculated values with the allele distribution model. Also, while examining the basis for the difference in calculated h, I noticed that conventional methods use the allele frequency (AF) of a variant in the general population to calculate the genetic variance of that variant. However, this implicitly assumes that the unaffected are included among the phenotypes of the disease - an assumption that is inconsistent with case-control studies in which unaffected individuals belong to the control (unaffected) group. Therefore, I modified the allele distribution model by using the AF in the patient population. Consequently, the h of rare variants calculated with the modified allele distribution model was quite high. Recalculating h of several rare variants reported in the literature with the modified allele distribution model yielded results were 3.2 - 53.7 times higher than the h calculated with the original allele distribution model. These results suggest that the contribution of rare variants to heritability of a disease has been considerably underestimated.