Prediction of complex phenotypes using theDrosophilametabolome

Abstract

Understanding the genotype -phenotype map and how variation at different levels of biological organization are associated are central topics in modern biology. Fast developments in sequencing technologies and other molecular omic tools enable researchers to obtain detailed information on variation at DNA level and on intermediate endophenotypes, such as RNA, proteins, metabolites. This facilitates our understanding of the link between genotypes and molecular and functional organismal phenotypes. Here, we use the Drosophila Genetic Reference Panel and nuclear magnetic resonance (NMR) metabolomics to investigate the ability of the metabolome to predict organismal phenotypes. We do this by performing NMR metabolomics on four replicate pools of male flies from each of 170 isogenic lines. We show first that metabolite profiles are variable among the investigated lines and that this variation is highly heritable. Second, we identify genes associated with metabolome variation. Third, we show that for four of five traits related to behaviour and stress resistance the metabolome predicts phenotypes more accurately than genomic data, and that the metabolites driving the increased accuracy was trait specific. Our comprehensive characterization of population-scale diversity of metabolomes and its genetic basis illustrates that metabolites have large potential as predictors of organismal phenotypes. This finding has strong applied importance e.g. in human medicine and animal and plant breeding.1 is a design matrix linking the genomic ( 1 ) and metabolomic ( 1 ) effects to the phenotypes. The random genomic effects are 1~( , [1,1] ! " ), and the metabolomic effects are 1~( , [1,1] 5 " ), where is the additive genomic relationship matrix as specified previously, and is the metabolomic relationship matrix.The metabolomic relationship matrix was computed as = ′ NMR ⁄ , where is a × NMR matrix of *77 18* T 6 , N 6 ), and was compared (using paired t-test corrected for multiple testing by a false discovery rate (FDR) of <0.05) within and across 9/ clusters to identify the NMR features resulting in the largest prediction accuracy. We only considered Drosophila melanogaster Genetic Reference Panel lines.

Metabolomic Spectra For Phenotypic Prediction of Malting Quality In Spring Barley

“…The performance of GBLUP and MBLUP was investigated in Drosophila, where the prediction accuracy for two behavioral traits was below 0.1 when based on GBLUP and then increased to above 0.4 when using MBLUP. Such an increase have also been found for two environmental stress resistance traits in Drosophila 23 . In the plant field, metabolic information was introduced into prediction of complex traits by Riedelsheimer, et al 33 , where the authors presented a complementary approach to exploit large-scale genomic and metabolic information in hybrid testcrosses.…”

“…With this information, it is worthwhile to investigate the role of MFs involved in the prediction of phenotypes for MQ traits. PLSR is a popular method used in the studies of metabolic profiles 22 , and using metabolomic BLUP (MBLUP) model gave better prediction accuracies than the BLUP model using genomic information for four of five quantitative traits investigated 23 .…”

Improving Genomic Prediction for Seed Quality Traits in Oat (Avena sativa L.) Using Trait-Specific Relationship Matrices

“…Omics technologies provide an easy and effective way to measure thousands of endophenotypes in large mapping populations. Many research groups are using these approaches to improve prediction for complex traits (Guo et al, 2016 ; Westhues et al, 2017 ; Rincent et al, 2018 ; Schrag et al, 2018 ; Li et al, 2019 ; Xiang et al, 2019 ; Rohde et al, 2020 ; Zhou et al, 2020 ). While several studies have reported improvements in prediction accuracies when these data were used to create relationship matrices, the results are often mixed and inconsistent (Guo et al, 2016 ; Schrag et al, 2018 ; Zhou et al, 2020 ).…”

“…Using these frameworks, Morgante et al (2020) showed that BLUP models that included relationship matrices derived from transciptome data, as well as transcriptome and genome-wide marker data improved prediction accuracies compared to models that used only genome-wide markers. Several other studies have reported similar improvements in prediction accuracies when omics-based kernels are used for prediction, suggesting that these omics-based kernels capture some component of the phenotype that is not explained by genome-wide markers (environmental or non-additive genetic variance) (Westhues et al, 2017;Rincent et al, 2018;Schrag et al, 2018;Krause et al, 2019;Li et al, 2019;Rohde et al, 2020;Zhou et al, 2020). Despite these promising studies, these improv2gfgements seem to be dependent on the trait, methodologies and datatype (Guo et al, 2016;Schrag et al, 2018;Zhou et al, 2020).…”

Leveraging Multiple Layers of Data To Predict Drosophila Complex Traits

“…While models including expression data could capture a greater amount of variance, their predictive ability was generally similar to GBLUP (Li et al 2019). Even more recently, two studies in Drosophila showed that using metabolites to predict a few complex traits can provide higher accuracy than genotypes (Zhou et al 2020;Rohde et al 2020).…”