Ascertainment Bias in the Genomic Test of Positive Selection on Regulatory Sequences.

Evolution of gene expression mediated by cis-regulatory changes is thought to be an important contributor to organismal adaptation, but identifying adaptive cis-regulatory changes is challenging due to the difficulty in knowing the expectation under no positive selection. A new approach for detecting positive selection on transcription factor binding sites (TFBSs) was recently developed, thanks to the application of machine learning in predicting transcription factor (TF) binding affinities of DNA sequences. Given a TFBS sequence from a focal species and the corresponding inferred ancestral sequence that differs from the former at n sites, one can predict the TF-binding affinities of many n-step mutational neighbors of the ancestral sequence and obtain a null distribution of the derived binding affinity, which allows testing whether the binding affinity of the real derived sequence deviates significantly from the null distribution. Applying this test genomically to all experimentally identified binding sites of 3 TFs in humans, a recent study reported positive selection for elevated binding affinities of TFBSs. Here, we show that this genomic test suffers from an ascertainment bias because, even in the absence of positive selection for strengthened binding, the binding affinities of known human TFBSs are more likely to have increased than decreased in evolution. We demonstrate by computer simulation that this bias inflates the false positive rate of the selection test. We propose several methods to mitigate the ascertainment bias and show that almost all previously reported positive selection signals disappear when these methods are applied.

On the Decoupling of Evolutionary Changes in mRNA and Protein Levels.

Variation in gene expression across lineages is thought to explain much of the observed phenotypic variation and adaptation. The protein is closer to the target of natural selection but gene expression is typically measured as the amount of mRNA. The broad assumption that mRNA levels are good proxies for protein levels has been undermined by a number of studies reporting moderate or weak correlations between the two measures across species. One biological explanation for this discrepancy is that there has been compensatory evolution between the mRNA level and regulation of translation. However, we do not understand the evolutionary conditions necessary for this to occur nor the expected strength of the correlation between mRNA and protein levels. Here, we develop a theoretical model for the coevolution of mRNA and protein levels and investigate the dynamics of the model over time. We find that compensatory evolution is widespread when there is stabilizing selection on the protein level; this observation held true across a variety of regulatory pathways. When the protein level is under directional selection, the mRNA level of a gene and the translation rate of the same gene were negatively correlated across lineages but positively correlated across genes. These findings help explain results from comparative studies of gene expression and potentially enable researchers to disentangle biological and statistical hypotheses for the mismatch between transcriptomic and proteomic data.

Genetic and selective constraints on the optimization of gene product diversity.

RNA and protein expressed from the same gene can have diverse isoforms due to various post-transcriptional and post-translational modifications. For the vast majority of alternative isoforms, It is unknown whether they are adaptive or simply biological noise. As we cannot experimentally probe the function of each isoform, we can ask whether the distribution of isoforms across genes and across species is consistent with expectations from different evolutionary processes. However, there is currently no theoretical framework that can generate such predictions. To address this, we developed a mathematical model where isoform abundances are determined collectively by cis-acting loci, trans-acting factors, gene expression levels, and isoform decay rates to predict isoform abundance distributions across species and genes in the face of mutation, genetic drift, and selection. We found that factors beyond selection, such as effective population size and the number of cis-acting loci, significantly influence evolutionary outcomes. Notably, suboptimal phenotypes are more likely to evolve when the population is small and/or when the number of cis-loci is large. We also explored scenarios where modification processes have both beneficial and detrimental effects, revealing a non-monotonic relationship between effective population size and optimization, demonstrating how opposing selection pressures on cis- and trans-acting loci can constrain the optimization of gene product diversity. As a demonstration of the power of our theory, we compared the expected distribution of A-to-I RNA editing levels in coleoids and found this to be largely consistent with non-adaptive explanations.

Detecting natural selection in trait-trait coevolution.

No phenotypic trait evolves independently of all other traits, but the cause of trait-trait coevolution is poorly understood. While the coevolution could arise simply from pleiotropic mutations that simultaneously affect the traits concerned, it could also result from multivariate natural selection favoring certain trait relationships. To gain a general mechanistic understanding of trait-trait coevolution, we examine the evolution of 220 cell morphology traits across 16 natural strains of the yeast Saccharomyces cerevisiae and the evolution of 24 wing morphology traits across 110 fly species of the family Drosophilidae, along with the variations of these traits among gene deletion or mutation accumulation lines (a.k.a. mutants). For numerous trait pairs, the phenotypic correlation among evolutionary lineages differs significantly from that among mutants. Specifically, we find hundreds of cases where the evolutionary correlation between traits is strengthened or reversed relative to the mutational correlation, which, according to our population genetic simulation, is likely caused by multivariate selection. Furthermore, we detect selection for enhanced modularity of the yeast traits analyzed. Together, these results demonstrate that trait-trait coevolution is shaped by natural selection and suggest that the pleiotropic structure of mutation is not optimal. Because the morphological traits analyzed here are chosen largely because of their measurability and thereby are not expected to be biased with regard to natural selection, our conclusion is likely general.

Ascertainment bias in the genomic test of positive selection on regulatory sequences.

Evolution of gene expression mediated by cis-regulatory changes is thought to be an important contributor to organismal adaptation, but identifying adaptive cis-regulatory changes is challenging due to the difficulty in knowing the expectation under no positive selection. A new approach for detecting positive selection on transcription factor binding sites (TFBSs) was recently developed, thanks to the application of machine learning in predicting transcription factor (TF) binding affinities of DNA sequences. Given a TFBS sequence from a focal species and the corresponding inferred ancestral sequence that differs from the former at n sites, one can predict the TF binding affinities of many n-step mutational neighbors of the ancestral sequence and obtain a null distribution of the derived binding affinity, which allows testing whether the binding affinity of the real derived sequence deviates significantly from the null distribution. Applying this test genomically to all experimentally identified binding sites of three TFs in humans, a recent study reported positive selection for elevated binding affinities of TFBSs. Here we show that this genomic test suffers from an ascertainment bias because, even in the absence of positive selection for strengthened binding, the binding affinities of known human TFBSs are more likely to have increased than decreased in evolution. We demonstrate by computer simulation that this bias inflates the false positive rate of the selection test. We propose several methods to mitigate the ascertainment bias and show that almost all previously reported positive selection signals disappear when these methods are applied.

Decoupling of evolutionary changes in mRNA and protein levels.

Variation in gene expression across lineages is thought to explain much of the observed phenotypic variation and adaptation. The protein is closer to the target of natural selection but gene expression is typically measured as the amount of mRNA. The broad assumption that mRNA levels are good proxies for protein levels has been undermined by a number of studies reporting moderate or weak correlations between the two measures across species. One biological explanation for this discrepancy is that there has been compensatory evolution between the mRNA level and regulation of translation. However, we do not understand the evolutionary conditions necessary for this to occur nor the expected strength of the correlation between mRNA and protein levels. Here we develop a theoretical model for the coevolution of mRNA and protein levels and investigate the dynamics of the model over time. We find that compensatory evolution is widespread when there is stabilizing selection on the protein level, which is true across a variety of regulatory pathways. When the protein level is under directional selection, the mRNA level of a gene and its translation rate of the same gene were negatively correlated across lineages but positively correlated across genes. These findings help explain results from comparative studies of gene expression and potentially enable researchers to disentangle biological and statistical hypotheses for the mismatch between transcriptomic and proteomic studies.

Evaluating the Performance of Widely Used Phylogenetic Models for Gene Expression Evolution.

Phylogenetic comparative methods are increasingly used to test hypotheses about the evolutionary processes that drive divergence in gene expression among species. However, it is unknown whether the distributional assumptions of phylogenetic models designed for quantitative phenotypic traits are realistic for expression data and importantly, the reliability of conclusions of phylogenetic comparative studies of gene expression may depend on whether the data is well described by the chosen model. To evaluate this, we first fit several phylogenetic models of trait evolution to 8 previously published comparative expression datasets, comprising a total of 54,774 genes with 145,927 unique gene-tissue combinations. Using a previously developed approach, we then assessed how well the best model of the set described the data in an absolute (not just relative) sense. First, we find that Ornstein-Uhlenbeck models, in which expression values are constrained around an optimum, were the preferred models for 66% of gene-tissue combinations. Second, we find that for 61% of gene-tissue combinations, the best-fit model of the set was found to perform well; the rest were found to be performing poorly by at least one of the test statistics we examined. Third, we find that when simple models do not perform well, this appears to be typically a consequence of failing to fully account for heterogeneity in the rate of the evolution. We advocate that assessment of model performance should become a routine component of phylogenetic comparative expression studies; doing so can improve the reliability of inferences and inspire the development of novel models.

Evaluating the Performance of Widely Used Phylogenetic Models for Gene Expression Evolution.

Phylogenetic comparative methods are increasingly used to test hypotheses about the evolutionary processes that drive divergence in gene expression among species. However, it is unknown whether the distributional assumptions of phylogenetic models designed for quantitative phenotypic traits are realistic for expression data and importantly, the reliability of conclusions of phylogenetic comparative studies of gene expression may depend on whether the data is well-described by the chosen model. To evaluate this, we first fit several phylogenetic models of trait evolution to 8 previously published comparative expression datasets, comprising a total of 54,774 genes with 145,927 unique gene-tissue combinations. Using a previously developed approach, we then assessed how well the best model of the set described the data in an absolute (not just relative) sense. First, we find that Ornstein-Uhlenbeck models, in which expression values are constrained around an optimum, were the preferred model for 66% of gene-tissue combinations. Second, we find that for 61% of gene-tissue combinations, the best fit model of the set was found to perform well; the rest were found to be performing poorly by at least one of the test statistics we examined. Third, we find that when simple models do not perform well, this appears to be typically a consequence of failing to fully account for heterogeneity in the rate of the evolution. We advocate that assessment of model performance should become a routine component of phylogenetic comparative expression studies; doing so can improve the reliability of inferences and inspire the development of novel models.

Fly wing evolution explained by a neutral model with mutational pleiotropy.

To what extent the speed of mutational production of phenotypic variation determines the rate of long-term phenotypic evolution is a central question. Houle et al. recently addressed this question by studying the mutational variances, additive genetic variances, and macroevolution of locations of vein intersections on fly wings, reporting very slow phenotypic evolution relative to the rates of mutational input, high phylogenetic signals, and a strong, linear relationship between the mutational variance of a trait and its rate of evolution. Houle et al. found no existing model of phenotypic evolution to be consistent with all these observations, and proposed the improbable scenario of equal influence of mutational pleiotropy on all traits. Here, we demonstrate that the purported linear relationship between mutational variance and evolutionary divergence is artifactual. We further show that the data are explainable by a simple model in which the wing traits are effectively neutral at least within a range of phenotypic values but their evolutionary rates are differentially reduced because mutations affecting these traits are purged owing to their different pleiotropic effects on other traits that are under stabilizing selection. Thus, the evolutionary patterns of fly wing morphologies are explainable under the existing theoretical framework of phenotypic evolution.

The preponderance of nonsynonymous A-to-I RNA editing in coleoids is nonadaptive.

A-to-I editing enzymatically converts the base adenosine (A) in RNA molecules to inosine (I), which is recognized as guanine (G) in translation. Exceptionally abundant A-to-I editing was recently discovered in the neural tissues of coleoids (octopuses, squids, and cuttlefishes), with a greater fraction of nonsynonymous sites than synonymous sites subject to high levels of editing. Although this phenomenon is thought to indicate widespread adaptive editing, its potential advantage is unknown. Here we propose an alternative, nonadaptive explanation. Specifically, increasing the cellular editing activity permits some otherwise harmful G-to-A nonsynonymous substitutions, because the As are edited to Is at sufficiently high levels. These high editing levels are constrained upon substitutions, resulting in the predominance of nonsynonymous editing at highly edited sites. Our evidence for this explanation suggests that the prevalent nonsynonymous editing in coleoids is generally nonadaptive, as in species with much lower editing activities.