Quantifying the fraction of new mutations that are recessive lethal

Abstract

The presence and impact of recessive lethal mutations has been widely documented in diploid outcrossing species. However, precise estimates in different species of the proportion of mutations that are recessive lethal remain limited. Here, we attempt to quantify the fraction of new mutations that are recessive lethal using Fit∂a∂i, a commonly-used method for inferring the distribution of fitness effects (DFE) using the site frequency spectrum. Using simulations, we demonstrate that Fit∂a∂i cannot accurately estimate the fraction of recessive lethal mutations, as expected given that Fit∂a∂i assumes that all mutations are additive by default. Consistent with the idea that mis-specification of the dominance model can explain this performance, we find that Fit∂a∂i can accurately infer the fraction of additive lethal mutations. Moreover, we demonstrate that in both additive and recessive cases, inference of the deleterious non-lethal portion of the DFE is minimally impacted by a small proportion (<10%) of lethal mutations. Finally, as an alternative approach to estimate the proportion of mutations that are recessive lethal, we employ models of mutation-selection-drift balance using existing genomic parameters and segregating recessive lethals estimates for humans and Drosophila melanogaster. In both species, we find that the segregating recessive lethal load can be explained by a very small fraction (<1%) of new nonsynonymous mutations being recessive lethal. Our results refute recent assertions of a much higher recessive lethal mutation fraction (4-5%), while highlighting the need for additional information on the joint distribution of selection and dominance coefficients.

Using computational simulations to quantify genetic load and predict extinction risk

“…Overall, this analysis demonstrates that sequence-based DFE estimates can explain empirical patterns of the inbreeding load when making slight adjustments to account for their shortcomings in estimating the proportion of recessive lethal mutations (80). Thus, these DFEs remain preferable for modelling deleterious variation in coding regions in that they account for the impacts of both weakly and strongly deleterious variation.…”

“…Although this value exceeds empirical estimates in humans (2B=1.4) and captive mammals (2B=3.1), this result is expected given that these estimates are based on juvenile survival and may therefore be underestimates (90, 91). Overall, this analysis demonstrates that sequence-based DFE estimates can explain empirical patterns of the inbreeding load when making slight adjustments to account for their shortcomings in estimating the proportion of recessive lethal mutations (80). Thus, these DFEs remain preferable for modelling deleterious variation in coding regions in that they account for the impacts of both weakly and strongly deleterious variation.…”

“…Finally, one important limitation of sequence-based approaches is that they typically assume that all mutations have additive effects on fitness, given that information on the distribution of dominance coefficients is very limited (though see (61)). Consequently, sequence-based DFE approaches may not be well powered for estimating the relatively small portion of the DFE that is highly recessive and strongly deleterious (80).…”

“…The observation that sequence-based DFEs contain a much smaller proportion of highly deleterious mutations compared to experimental-based DFEs is not surprising, given that experimental studies are known to be biased towards detecting strongly deleterious mutations (65, 75) and sequence-based studies may be somewhat underpowered to detect such variation (80). However, if experimental-based studies are presumed to estimate the full DFE for nonsynonymous mutations, they should nevertheless make predictions that are consistent with patterns of nonsynonymous variation in genomic sequencing datasets.…”

“…Given the shortcomings of these existing models, we propose a new model based on an analysis of the DFE in humans under non-additive model (93) as well as an estimation of the recessive lethal portion of the DFE (80). In brief, this ‘best available’ model assumes a somewhat less recessive dominance distribution compared to the Henn et al (25) distribution assumed by Kyriazis et al (59) model and is augmented with a small proportion (0.3%) of recessive lethal mutations (see Supplemental Appendix 3 for details).…”