Human data
To validate our intercept method, we ran it on 26 human samples from the 1000 Genomes project. Assuming a genome size of 3.5 Gbp, these data corresponded to a genome coverages ranging from 0.28x to 1.8x (median of 0.65x). The alignment rates of reads to a human mitochondrial reference had a bimodal distribution with 4 samples showing extremely low rates close to 0.05%. The greatest intercept (dashed line in Fig. 2) generated an estimate for the NUMT composition of the nuclear genome of 0.014% (95% confidence interval: 0.011%-0.018%). The estimate based on the lowest proportion of mapped reads (“mapping depth estimate”, solid line) was 0.054%.