Through imputation, we were able to take advantage of the moderate-depth sequences we generated, and the wealth of genetic information contained in them: over 27 million autosomal variants, many of which are rare. The first step was to phase the array genotypes for the full sample of 7,278 genotyped individuals in the MCTFR (excluding ungenotyped MZ co-twins). We used the full MCTFR sample in order to improve phasing accuracy with SHAPEIT (Delaneau, Zagury, & Marchini, 2013) by using as much family information as possible. By using the full sample for this purpose, we retained all available family information, thereby maximizing phasing accuracy.
