Genic constraint against nonsynonymous variation across the mouse genome

Abstract

Background: Selective constraint, the depletion of variation due to negative selection, provides insights into the functional impact of variants and disease mechanisms. However, its characterization in mice, the most commonly used mammalian model, remains limited. This study aims to quantify mouse gene constraint using a new metric called the nonsynonymous observed expected ratio (NOER) and investigate its relationship with gene function.

Results: NOER was calculated using whole-genome sequencing data from wild mouse populations (Mus musculus sp and Mus spretus). Positive correlations were observed between mouse gene constraint and the number of associated knockout phenotypes, indicating stronger constraint on pleiotropic genes. Furthermore, mouse gene constraint showed a positive correlation with the number of pathogenic variant sites in their human orthologues, supporting the relevance of mouse models in studying human disease variants.

Conclusions: NOER provides a resource for assessing the fitness consequences of genetic variants in mouse genes and understanding the relationship between gene constraint and function. The study's findings highlight the importance of pleiotropy in selective constraint and support the utility of mouse models in investigating human disease variants. Further research with larger sample sizes can refine constraint estimates in mice and enable more comprehensive comparisons of constraint between mouse and human orthologues.

Keywords: Mouse models; Negative selection; Selective constraint; Synonymous and nonsynonymous mutation.

Fig. 1

– (A) Distributions of gene constraint by lethal (L), sub-viable (SV), viable ‘with phenotype’ (VP), and viable ‘no phenotype’ (VN) knockout groupings. Constraint is quantified as the nonsynonymous observed expected ratio (NOER) and dN/dS, with a higher percentile rank indicating a greater degree of constraint. The dashed line indicates the median score for all genes under consideration. The grey areas indicate the 10% most and 10% least constrained genes. (B and C) Odds ratios and 95% confidence intervals for the 10% most constrained (n = 557), and 10% least constrained (n = 371) genes across each knockout group. An odds ratio greater than one indicates enrichment. The least constrained genes are enriched for genes with no associated phenotype deviation (B), and the most constrained genes are enriched for lethal and sub-viable knockout phenotypes (C). (D) ROC curves highlight the accuracy of NOER and dN/dS to predict knockout lethality or sub-viability. We fit two simple logistic regressions to predict knockout lethality or sub-viability, one with NOER and the other with dN/dS as the explanatory variable. NOER has greater predictive accuracy than dN/dS (0.65 compared with 0.59). (E) Scatter plot showing the positive Spearman’s correlation between constraint (NOER and dN/dS) and pleiotropy. Pleiotropy is defined here by hit rate: the number of top-level mouse phenotype ontology terms associated with the knock-out divided by the potential number of top-level terms based on the conducted phenotyping tests. (F) The 10% most constrained mouse genes are significantly more pleiotropic than the 10% least constrained mouse genes (p = 4.4e-12). Dashed lines represent the means for each group

Fig. 2

– (A and B) Mouse gene constraint is positively correlated with the number of known pathogenic variants in their human orthologues and negatively correlated with the number of known benign variants. 15,501 pathogenic and 38,423 benign SNV sites were considered from the ClinVar database across 7,016 human genes with a mouse one-to-one orthologue. Constraint is quantified as the NOER, with a greater percentile rank indicating a greater degree of constraint. Rank correlations were calculated between the mean number of pathogenic and benign variants per kb in the human orthologue and the NOER percentile bin. (C) Odds ratios and 95% confidence intervals for the enrichment of genes containing 3 or more pathogenic variants in their human orthologue amongst the 10% most constrained mouse genes defined by NOER (n = 1,474) and dN/dS (n = 1,076). The most constrained mouse genes are enriched for human orthologues with 3 or more pathogenic variants and depleted of human orthologues with 3 or more benign variants. (D) Spearman’s Rank correlation matrix of constraint scores for human and mouse one-to-one orthologues. Mouse gene constraint is calculated as the nonsynonymous observed expected ratio (NOER), and dN/dS. Human gene constraint scores include the missense Z-score, loss-of-function observed/expected upper bound fraction (LOEUF), and the probability of loss-of-function intolerance (pLI). NOER is significantly (Spearman’s rho, p < 0.0001) correlated with metrics of human gene constraint and mouse dN/dS