Comparing statistical learning methods for complex trait prediction from gene expression

Accurate prediction of complex traits is an important task in quantitative genetics. Genotypes have been used for trait prediction using a variety of methods such as mixed models, Bayesian methods, penalized regression methods, dimension reduction methods, and machine learning methods. Recent studies have shown that gene expression levels can produce higher prediction accuracy than genotypes. However, only a few prediction methods were tested in these studies. Thus, a comprehensive assessment of methods is needed to fully evaluate the potential of gene expression as a predictor of complex trait phenotypes. Here, we used data from the Drosophila Genetic Reference Panel (DGRP) to compare the ability of several existing statistical learning methods to predict starvation resistance and startle response from gene expression in the two sexes separately. The methods considered differ in assumptions about the distribution of gene effects-ranging from models that assume that every gene affects the trait to more sparse models-and their ability to capture gene-gene interactions. We also used functional annotation (i.e., Gene Ontology (GO)) as a source of biological information to inform prediction models. The results show that differences in prediction accuracy exist. For example, methods performing variable selection achieved higher prediction accuracy for starvation resistance in females, while they generally had lower accuracy for startle response in both sexes. Incorporating GO annotations further improved prediction accuracy for a few GO terms of biological significance. Biological significance extended to the genes underlying highly predictive GO terms. Notably, the Insulin-like Receptor (InR) was prevalent across methods and sexes for starvation resistance. For startle response, crumbs (crb) and imaginal disc growth factor 2 (Idgf2) were found for females and males, respectively. Our results confirmed the potential of transcriptomic prediction and highlighted the importance of selecting appropriate methods and strategies in order to achieve accurate predictions.

Pleiotropy, epistasis and the genetic architecture of quantitative traits

Pleiotropy (whereby one genetic polymorphism affects multiple traits) and epistasis (whereby non-linear interactions between genetic polymorphisms affect the same trait) are fundamental aspects of the genetic architecture of quantitative traits. Recent advances in the ability to characterize the effects of polymorphic variants on molecular and organismal phenotypes in human and model organism populations have revealed the prevalence of pleiotropy and unexpected shared molecular genetic bases among quantitative traits, including diseases. By contrast, epistasis is common between polymorphic loci associated with quantitative traits in model organisms, such that alleles at one locus have different effects in different genetic backgrounds, but is rarely observed for human quantitative traits and common diseases. Here, we review the concepts and recent inferences about pleiotropy and epistasis, and discuss factors that contribute to similarities and differences between the genetic architecture of quantitative traits in model organisms and humans.

An analysis of direct and indirect effects in Drosophila melanogaster undergoing a few cycles of experimental evolution for stress-related traits

The physiological mechanisms underpinning adaptations to starvation and cold stresses have been extensively studied in Drosophila, yet the understanding of correlated changes in stress-related and life-history traits, as well as the energetics of stress tolerance, still remains elusive. To answer the questions empirically in this context, we allowed D. melanogaster to evolve for either increased starvation or cold tolerance (24-generations / regime) in an experimental evolution system, and examined whether selection of either trait affects un-selected stress trait, as well as the impacts potential changes in life-history and mating success-related traits. Our results revealed remarkable changes in starvation/cold tolerance (up to 1.5-fold) as a direct effect of selection, while cold tolerance had been dramatically reduced (1.26-fold) in the starvation tolerant (ST) lines compared to control counterparts, although no such changes were evident in cold-tolerant (CT) lines. ST lines exhibited a higher level of body lipids and a reduced level of trehalose content, while CT lines accumulated a greater levels of body lipid and trehalose contents. Noticeably, we found that selection for starvation or cold tolerance positively correlates with larval development time, longevity, and copulation duration, indicating that these traits are among the most common targets of selection trajectories shaping stress tolerance. Altogether, this study highlights the complexity of mechanisms evolved in ST lines that contribute to enhanced starvation tolerance, but also negatively impact cold tolerance. Nevertheless, mechanisms foraging enhanced cold tolerance in CT lines appear not to target starvation tolerance. Moreover, the parallel changes in life history/mating success traits across stress regimes could indicate some generic pathways evolved in stressful environments, targeting life-history and mating success characteristics to optimize fitness.

Physiological and metabolomic consequences of reduced expression of the Drosophila brummer triglyceride Lipase

Disruption of lipolysis has widespread effects on intermediary metabolism and organismal phenotypes. Defects in lipolysis can be modeled in Drosophila melanogaster through genetic manipulations of brummer (bmm), which encodes a triglyceride lipase orthologous to mammalian Adipose Triglyceride Lipase. RNAi-mediated knock-down of bmm in all tissues or metabolic specific tissues results in reduced locomotor activity, altered sleep patterns and reduced lifespan. Metabolomic analysis on flies in which bmm is downregulated reveals a marked reduction in medium chain fatty acids, long chain saturated fatty acids and long chain monounsaturated and polyunsaturated fatty acids, and an increase in diacylglycerol levels. Elevated carbohydrate metabolites and tricarboxylic acid intermediates indicate that impairment of fatty acid mobilization as an energy source may result in upregulation of compensatory carbohydrate catabolism. bmm downregulation also results in elevated levels of serotonin and dopamine neurotransmitters, possibly accounting for the impairment of locomotor activity and sleep patterns. Physiological phenotypes and metabolomic changes upon reduction of bmm expression show extensive sexual dimorphism. Altered metabolic states in the Drosophila model are relevant for understanding human metabolic disorders, since pathways of intermediary metabolism are conserved across phyla.

A screen for sleep and starvation resistance identifies a wake-promoting role for the auxiliary channel unc79

The regulation of sleep and metabolism are highly interconnected, and dysregulation of sleep is linked to metabolic diseases that include obesity, diabetes, and heart disease. Furthermore, both acute and long-term changes in diet potently impact sleep duration and quality. To identify novel factors that modulate interactions between sleep and metabolic state, we performed a genetic screen for their roles in regulating sleep duration, starvation resistance, and starvation-dependent modulation of sleep. This screen identified a number of genes with potential roles in regulating sleep, metabolism, or both processes. One such gene encodes the auxiliary ion channel UNC79, which was implicated in both the regulation of sleep and starvation resistance. Genetic knockdown or mutation of unc79 results in flies with increased sleep duration, as well as increased starvation resistance. Previous findings have shown that unc79 is required in pacemaker for 24-hours circadian rhythms. Here, we find that unc79 functions in the mushroom body, but not pacemaker neurons, to regulate sleep duration and starvation resistance. Together, these findings reveal spatially localized separable functions of unc79 in the regulation of circadian behavior, sleep, and metabolic function.

Functional Diversification, Redundancy, and Epistasis among Paralogs of the Drosophila melanogaster Obp50a-d Gene Cluster

Large multigene families, such as the insect odorant-binding proteins (OBPs), are thought to arise through functional diversification after repeated gene duplications. Whereas many OBPs function in chemoreception, members of this family are also expressed in tissues outside chemosensory organs. Paralogs of the Obp50 gene cluster are expressed in metabolic and male reproductive tissues, but their functions and interrelationships remain unknown. Here, we report the genetic dissection of four members of the Obp50 cluster, which are in close physical proximity without intervening genes. We used CRISPR technology to excise the entire cluster while introducing a PhiC31 reintegration site to reinsert constructs in which different combinations of the constituent Obp genes were either intact or rendered inactive. We performed whole transcriptome sequencing and assessed sexually dimorphic changes in transcript abundances (transcriptional niches) associated with each gene-edited genotype. Using this approach, we were able to estimate redundancy, additivity, diversification, and epistasis among Obp50 paralogs. We analyzed the effects of gene editing of this cluster on organismal phenotypes and found a significant skewing of sex ratios attributable to Obp50a, and sex-specific effects on starvation stress resistance attributable to Obp50d. Thus, there is functional diversification within the Obp50 cluster with Obp50a contributing to development and Obp50d to stress resistance. The deletion-reinsertion approach we applied to the Obp50 cluster provides a general paradigm for the genetic dissection of paralogs of multigene families.

Regulatory Mechanisms of Metamorphic Neuronal Remodeling Revealed Through a Genome-Wide Modifier Screen in Drosophila melanogaster

During development, neuronal remodeling shapes neuronal connections to establish fully mature and functional nervous systems. Our previous studies have shown that the RNA-binding factor alan shepard (shep) is an important regulator of neuronal remodeling during metamorphosis in Drosophila melanogaster, and loss of shep leads to smaller soma size and fewer neurites in a stage-dependent manner. To shed light on the mechanisms by which shep regulates neuronal remodeling, we conducted a genetic modifier screen for suppressors of shep-dependent wing expansion defects and cellular morphological defects in a set of peptidergic neurons, the bursicon neurons, that promote posteclosion wing expansion. Out of 702 screened deficiencies that covered 86% of euchromatic genes, we isolated 24 deficiencies as candidate suppressors, and 12 of them at least partially suppressed morphological defects in shep mutant bursicon neurons. With RNA interference and mutant alleles of individual genes, we identified Daughters against dpp (Dad) and Olig family (Oli) as shep suppressor genes, and both of them restored the adult cellular morphology of shep-depleted bursicon neurons. Dad encodes an inhibitory Smad protein that inhibits bone morphogenetic protein (BMP) signaling, raising the possibility that shep interacted with BMP signaling through antagonism of Dad. By manipulating expression of the BMP receptor tkv, we found that activated BMP signaling was sufficient to rescue loss-of-shep phenotypes. These findings reveal mechanisms of shep regulation during neuronal development, and they highlight a novel genetic shep interaction with the BMP signaling pathway that controls morphogenesis in mature, terminally differentiated neurons during metamorphosis.

Genomic Prediction for Quantitative Traits Is Improved by Mapping Variants to Gene Ontology Categories in Drosophila melanogaster

Predicting individual quantitative trait phenotypes from high-resolution genomic polymorphism data is important for personalized medicine in humans, plant and animal breeding, and adaptive evolution. However, this is difficult for populations of unrelated individuals when the number of causal variants is low relative to the total number of polymorphisms and causal variants individually have small effects on the traits. We hypothesized that mapping molecular polymorphisms to genomic features such as genes and their gene ontology categories could increase the accuracy of genomic prediction models. We developed a genomic feature best linear unbiased prediction (GFBLUP) model that implements this strategy and applied it to three quantitative traits (startle response, starvation resistance, and chill coma recovery) in the unrelated, sequenced inbred lines of the Drosophila melanogaster Genetic Reference Panel. Our results indicate that subsetting markers based on genomic features increases the predictive ability relative to the standard genomic best linear unbiased prediction (GBLUP) model. Both models use all markers, but GFBLUP allows differential weighting of the individual genetic marker relationships, whereas GBLUP weighs the genetic marker relationships equally. Simulation studies show that it is possible to further increase the accuracy of genomic prediction for complex traits using this model, provided the genomic features are enriched for causal variants. Our GFBLUP model using prior information on genomic features enriched for causal variants can increase the accuracy of genomic predictions in populations of unrelated individuals and provides a formal statistical framework for leveraging and evaluating information across multiple experimental studies to provide novel insights into the genetic architecture of complex traits.

Multiple-Line Inference of Selection on Quantitative Traits

Trait differences between species may be attributable to natural selection. However, quantifying the strength of evidence for selection acting on a particular trait is a difficult task. Here we develop a population genetics test for selection acting on a quantitative trait that is based on multiple-line crosses. We show that using multiple lines increases both the power and the scope of selection inferences. First, a test based on three or more lines detects selection with strongly increased statistical significance, and we show explicitly how the sensitivity of the test depends on the number of lines. Second, a multiple-line test can distinguish between different lineage-specific selection scenarios. Our analytical results are complemented by extensive numerical simulations. We then apply the multiple-line test to QTL data on floral character traits in plant species of the Mimulus genus and on photoperiodic traits in different maize strains, where we find a signature of lineage-specific selection not seen in two-line tests.

Genetic Architecture of Micro-Environmental Plasticity in Drosophila melanogaster

Individuals of the same genotype do not have the same phenotype for quantitative traits when reared under common macro-environmental conditions, a phenomenon called micro-environmental plasticity. Genetic variation in micro-environmental plasticity is assumed in models of the evolution of phenotypic variance, and is important in applied breeding and personalized medicine. Here, we quantified genetic variation for micro-environmental plasticity for three quantitative traits in the inbred, sequenced lines of the Drosophila melanogaster Genetic Reference Panel. We found substantial genetic variation for micro-environmental plasticity for all traits, with broad sense heritabilities of the same magnitude or greater than those of trait means. Micro-environmental plasticity is not correlated with residual segregating variation, is trait-specific, and has genetic correlations with trait means ranging from zero to near unity. We identified several candidate genes associated with micro-environmental plasticity of startle response, including Drosophila Hsp90, setting the stage for future genetic dissection of this phenomenon.

Neuronal remodeling during metamorphosis is regulated by the alan shepard (shep) gene in Drosophila melanogaster

Peptidergic neurons are a group of neuronal cells that synthesize and secrete peptides to regulate a variety of biological processes. To identify genes controlling the development and function of peptidergic neurons, we conducted a screen of 545 splice-trap lines and identified 28 loci that drove expression in peptidergic neurons when crossed to a GFP reporter transgene. Among these lines, an insertion in the alan shepard (shep) gene drove expression specifically in most peptidergic neurons. shep transcripts and SHEP proteins were detected primarily and broadly in the central nervous system (CNS) in embryos, and this expression continued into the adult stage. Loss of shep resulted in late pupal lethality, reduced adult life span, wing expansion defects, uncoordinated adult locomotor activities, rejection of males by virgin females, and reduced neuropil area and reduced levels of multiple presynaptic markers throughout the adult CNS. Examination of the bursicon neurons in shep mutant pharate adults revealed smaller somata and fewer axonal branches and boutons, and all of these cellular phenotypes were fully rescued by expression of the most abundant wild-type shep isoform. In contrast to shep mutant animals at the pharate adult stage, shep mutant larvae displayed normal bursicon neuron morphologies. Similarly, shep mutant adults were uncoordinated and weak, while shep mutant larvae displayed largely, although not entirely, normal locomotor behavior. Thus, shep played an important role in the metamorphic development of many neurons.

The genetics of quantitative traits: challenges and prospects

A major challenge in current biology is to understand the genetic basis of variation for quantitative traits. We review the principles of quantitative trait locus mapping and summarize insights about the genetic architecture of quantitative traits that have been obtained over the past decades. We are currently in the midst of a genomic revolution, which enables us to incorporate genetic variation in transcript abundance and other intermediate molecular phenotypes into a quantitative trait locus mapping framework. This systems genetics approach enables us to understand the biology inside the 'black box' that lies between genotype and phenotype in terms of causal networks of interacting genes.

Quantitative trait loci for aggressive behavior in Drosophila melanogaster

Aggressive behavior is observed across animal taxa and is likely to be evolutionarily conserved. Although potentially advantageous, aggression can have social and health consequences in humans, and is a component of a number of psychiatric disorders. As a complex genetic trait, it is modulated by numerous quantitative trait loci (QTL) with allelic effects that can vary in direction and magnitude and that are sensitive to environmental perturbations. Assays to quantify aggressive behavior in Drosophila melanogaster have been developed, making this an ideal model system in which to dissect the genomic architecture underlying manifestation of and variation in aggressive behavior. Here, we map QTL affecting variation in aggression between two wild-type Drosophila strains. We identified a minimum of five QTL in a genomewide scan: two on chromosome 2 and three on chromosome 3. At least three and possibly all five of these QTL interact epistatically. We used deficiency complementation mapping to subdivide two linked, epistatically interacting QTL of large effect on chromosome 3 into at least six QTL, and complementation tests to mutations identified four candidate quantitative trait genes. Extensive epistasis poses a serious challenge for understanding the genetic basis of complex traits.

Quantitative genetic analysis of sleep in Drosophila melanogaster

Although intensively studied, the biological purpose of sleep is not known. To identify candidate genes affecting sleep, we assayed 136 isogenic P-element insertion lines of Drosophila melanogaster. Since sleep has been negatively correlated with energy reserves across taxa, we measured energy stores (whole-body protein, glycogen, and triglycerides) in these lines as well. Twenty-one insertions with known effects on physiology, development, and behavior affect 24-hr sleep time. Thirty-two candidate insertions significantly impact energy stores. Mutational genetic correlations among sleep parameters revealed that the genetic basis of the transition between sleep and waking states in males and females may be different. Furthermore, sleep bout number can be decoupled from waking activity in males, but not in females. Significant genetic correlations are present between sleep phenotypes and glycogen stores in males, while sleep phenotypes are correlated with triglycerides in females. Differences observed in male and female sleep behavior in flies may therefore be related to sex-specific differences in metabolic needs. Sleep thus emerges as a complex trait that exhibits extensive pleiotropy and sex specificity. The large mutational target that we observed implicates genes functioning in a variety of biological processes, suggesting that sleep may serve a number of different functions rather than a single purpose.

Quantitative trait loci for locomotor behavior in Drosophila melanogaster

Locomotion is an integral component of most animal behaviors and many human diseases and disorders are associated with locomotor deficits, but little is known about the genetic basis of natural variation in locomotor behavior. Locomotion is a complex trait, with variation attributable to the joint segregation of multiple interacting quantitative trait loci (QTL), with effects that are sensitive to the environment. We assessed variation in a component of locomotor behavior (locomotor reactivity) in a population of 98 recombinant inbred lines of Drosophila melanogaster and mapped four QTL affecting locomotor reactivity by linkage to polymorphic roo transposable element insertion sites. We used complementation tests of deficiencies to fine map these QTL to 12 chromosomal regions and complementation tests of mutations to identify 13 positional candidate genes affecting locomotor reactivity, including Dopa decarboxylase (Ddc), which catalyzes the final step in the synthesis of serotonin and dopamine. Linkage disequilibrium mapping in a population of 164 second chromosome substitution lines derived from a single natural population showed that polymorphisms at Ddc were associated with naturally occurring genetic variation in locomotor behavior. These data implicate variation in the synthesis of bioamines as a factor contributing to natural variation in locomotor reactivity.

Quantitative trait loci with age-specific effects on fecundity in Drosophila melanogaster

Life-history theory and evolutionary theories of aging assume the existence of alleles with age-specific effects on fitness. While various studies have documented age-related changes in the genetic contribution to variation in fitness components, we know very little about the underlying genetic architecture of such changes. We used a set of recombinant inbred lines to map and characterize the effects of quantitative trait loci (QTL) affecting fecundity of Drosophila melanogaster females at 1 and 4 weeks of age. We identified one QTL on the second chromosome and one or two QTL affecting fecundity on the third chromosome, but these QTL affected fecundity only at 1 week of age. There was more genetic variation for fecundity at 4 weeks of age than at 1 week of age and there was no genetic correlation between early and late-age fecundity. These results suggest that different loci contribute to the variation in fecundity as the organism ages. Our data provide support for the mutation accumulation theory of aging as applied to reproductive senescence. Comparing the results from this study with our previous work on life-span QTL, we also find evidence that antagonistic pleiotropy may contribute to the genetic basis of senescence in these lines as well.

The BDGP gene disruption project: single transposon insertions associated with 40% of Drosophila genes

The Berkeley Drosophila Genome Project (BDGP) strives to disrupt each Drosophila gene by the insertion of a single transposable element. As part of this effort, transposons in >30,000 fly strains were localized and analyzed relative to predicted Drosophila gene structures. Approximately 6300 lines that maximize genomic coverage were selected to be sent to the Bloomington Stock Center for public distribution, bringing the size of the BDGP gene disruption collection to 7140 lines. It now includes individual lines predicted to disrupt 5362 of the 13,666 currently annotated Drosophila genes (39%). Other lines contain an insertion at least 2 kb from others in the collection and likely mutate additional incompletely annotated or uncharacterized genes and chromosomal regulatory elements. The remaining strains contain insertions likely to disrupt alternative gene promoters or to allow gene misexpression. The expanded BDGP gene disruption collection provides a public resource that will facilitate the application of Drosophila genetics to diverse biological problems. Finally, the project reveals new insight into how transposons interact with a eukaryotic genome and helps define optimal strategies for using insertional mutagenesis as a genomic tool.

Genetics and genomics of Drosophila mating behavior

The first steps of animal speciation are thought to be the development of sexual isolating mechanisms. In contrast to recent progress in understanding the genetic basis of postzygotic isolating mechanisms, little is known about the genetic architecture of sexual isolation. Here, we have subjected Drosophila melanogaster to 29 generations of replicated divergent artificial selection for mating speed. The phenotypic response to selection was highly asymmetrical in the direction of reduced mating speed, with estimates of realized heritability averaging 7%. The selection response was largely attributable to a reduction in female receptivity. We assessed the whole genome transcriptional response to selection for mating speed using Affymetrix GeneChips and a rigorous statistical analysis. Remarkably, >3,700 probe sets (21% of the array elements) exhibited a divergence in message levels between the Fast and Slow replicate lines. Genes with altered transcriptional abundance in response to selection fell into many different biological process and molecular function Gene Ontology categories, indicating substantial pleiotropy for this complex behavior. Future functional studies are necessary to test the extent to which transcript profiling of divergent selection lines accurately predicts genes that directly affect the selected trait.

Mapping multiple Quantitative Trait Loci by Bayesian classification

We developed a classification approach to multiple quantitative trait loci (QTL) mapping built upon a Bayesian framework that incorporates the important prior information that most genotypic markers are not cotransmitted with a QTL or their QTL effects are negligible. The genetic effect of each marker is modeled using a three-component mixture prior with a class for markers having negligible effects and separate classes for markers having positive or negative effects on the trait. The posterior probability of a marker's classification provides a natural statistic for evaluating credibility of identified QTL. This approach performs well, especially with a large number of markers but a relatively small sample size. A heat map to visualize the results is proposed so as to allow investigators to be more or less conservative when identifying QTL. We validated the method using a well-characterized data set for barley heading values from the North American Barley Genome Mapping Project. Application of the method to a new data set revealed sex-specific QTL underlying differences in glucose-6-phosphate dehydrogenase enzyme activity between two Drosophila species. A simulation study demonstrated the power of this approach across levels of trait heritability and when marker data were sparse.