Genomic DNA was extracted from each DO mouse (n = 192 total samples) and genotyped at 57,973 single nucleotide polymorphisms (SNPs) on the Mega-MUGA platform (Geneseek)44 (link). A total of 177 out of 192 samples passed SNP quality control metrics. For these samples, founder haplotypes were inferred from SNP probe intensities using a hidden Markov model implemented in the DOQTL R package27 (link),45 (link), and then used to interpolate a grid of 64,000 evenly-spaced genetic intervals. In addition, founder haplotypes were independently inferred from the RNA-seq data by genotyping by RNA-seq (GBRS) protocol (see next section) and interpolated to the same 64,000 interval grid.
For each sample, we verified that the haplotype reconstructions agreed between the DNA Mega-MUGA and GBRS reconstructions by calculating the Pearson correlation between each pair of samples. When a Mega-MUGA sample had a correlation below 0.4 with the same sample ID in the RNA-seq data, we assumed that this sample was mismatched. We searched the RNA-seq data for the correct match to the Mega-MUGA sample by looking for another sample that was more highly correlated. If we found an RNA-seq sample with a correlation >0.4 that was not assigned to another sample, we matched it with the Mega-MUGA sample. When a sample was removed from the Mega-MUGA data for technical reasons, we used the GBRS haplotype reconstructions (samples F326, F328, F362, F363, F368, M377, M388, M392, M393, M394, M404, M408, M411, M419 and M425).
For each sample, we verified that the haplotype reconstructions agreed between the DNA Mega-MUGA and GBRS reconstructions by calculating the Pearson correlation between each pair of samples. When a Mega-MUGA sample had a correlation below 0.4 with the same sample ID in the RNA-seq data, we assumed that this sample was mismatched. We searched the RNA-seq data for the correct match to the Mega-MUGA sample by looking for another sample that was more highly correlated. If we found an RNA-seq sample with a correlation >0.4 that was not assigned to another sample, we matched it with the Mega-MUGA sample. When a sample was removed from the Mega-MUGA data for technical reasons, we used the GBRS haplotype reconstructions (samples F326, F328, F362, F363, F368, M377, M388, M392, M393, M394, M404, M408, M411, M419 and M425).