Assessing variant conservation using genome-wide phyloP scores

PhyloP scores for each base were downloaded from the UCSC genome browser (http://hgdownload.cse.ucsc.edu/goldenpath/hg38/phyloP100way/). 1-indexed bigwig files were converted to bed file format using the wig2bed tool from bedops (version 2.4.36; https://bedops.readthedocs.io/en/latest/index.html). These base-level annotations were matched to each uORF base and used to determine the proportion of bases that were significantly conserved (proportion of bases with phyloP score >2). Possible inframe stop-codon creating positions were identified based on mapped reading frames for each uORF. These sites were extracted and further categorized by whether a SNV could create a UGA, UAG, or UAA stop codon. Some positions could be mutated to either a UGA, UAG, or UAA codon and these were considered separately from potential UGA, UAG, or UAA-creating positions. We have included all potential stop-introducing positions as Supplementary Data ¹. As a control we used phyloP scores for genomic positions with the potential to create non-uORF UGA, UAG, or UAA trinucleotides by mutation, but matched by distance to CDS in 10-base pair windows.

Start-disrupting genomic positions were annotated as those mutating the second or third position in the first codon of each translated uORF. Conservation based on phyloP scores were assessed for start-disrupting positions similar to potential stop-introducing positions. As a control we compared phyloP scores for uORF start-disrupting positions to out-of-frame start-disrupting positions within annotated uORFs, and a set of NTG start-disrupting variants that were not part of translated uORFs but matched by distance to the CDS as determined by 10-bp windows.

P values were determined by sampling with replacement from each set of variants 10,000 times and re-calculating the proportion of significantly conserved bases (phyloP score >2). The distribution of the fraction of conserved base positions were then compared against different sets of variants, and the P value was defined as the fraction of samples where one group was higher than the other.