Determining the effect of strong purifying selection on coding sequences

Our results had revealed significant biases in mutation probability in relation to gene bodies. Because we had found that mutations were significantly higher upstream of genes and significantly lower within gene bodies in five independent datasets, we considered the possibility that this overwhelming bias was the result of extremely strong purifying selection on de novo mutations (that is, removal of lethal mutations before they could be detected by us). We therefore simulated 10,000 random mutations across the TAIR10 genome. If mutations fell within coding regions, we randomly assigned them to be removed by selection (that is, dominant lethal). For this, we explored three levels of selection: s = 0.01 where 1% of mutations were removed (that is, had lethal effects), s = 0.1 where 10% of mutations were removed, s = 0.2 where 20% of mutations were removed, or s = 0.3 where 30% of mutations were removed. While s = 0.3 represents an exceptionally and unexpectedly high level of selection, especially in soma, evidenced by empirical estimates of the extent of gene essentiality in A. thaliana, this served as a positive control for observing the effects of extraordinarily strong selection on the expected distribution of mutations in a random mutation model.