iKIR score interaction terms with protective HLA class II drivers adjusted by relatedness in the cohort after exclusion of HLA class I drivers.

<p>After removal of class I drivers from the cohort (remaining cohort: N <b>=</b> 5,420), we modelled risk of T1D for each protective genotype and included iKIR score in the model as a continuous variable interacting with the HLA class II genotype (). To account for relatedness between individuals, we used a generalized mixed model including a genetic relatedness matrix as a random factor. The coefficient for the interaction term was in the expected direction for all genotypes and significant for 5 frequent genotypes. By comparison with <b><a href="http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1011456#pgen.1011456.t002" target="_blank">Table 2</a></b> (same analysis but without the genetic relatedness matrix) it can be seen that adjusting for cryptic relatedness had little impact (all coefficients and p values very similar, <b><a href="http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1011456#pgen.1011456.g004" target="_blank">Fig 4</a></b>).</p> <p>(PDF)</p>