Comments from curator
No phenotype data: this study was rejected because no new phenotype was measured on DGRP lines.This is a genomic prediction study. It was first posted on bioRxiv (
10.1101/824896, 2019) and published in
G3 in 2020 (
10.1534/g3.120.401847;
PMC version). The authors compared how well three kinds of data predict three traits in females and males, using 30 replicates of fivefold cross-validation:
- DNA variants (GBLUP; 1,908,995 variants with MAF > 0.05 and call rate > 0.8).
- Whole-fly gene expression (TBLUP; RNA-seq of 200 DGRP lines from Everett et al., 2020).
- Combinations of the two, with or without Gene Ontology annotation.
The three traits are the line means of starvation resistance, chill coma recovery time and startle response published by
Mackay et al., 2012. Of the lines with expression data, 198 had starvation resistance, 172 chill coma recovery and 199 startle response. The analysis code adjusts these means for
Wolbachia infection and the five major inversions. The traits are already in DGRPool:
- Study 21 has the line means (mn_StarvationRes, mn_ChillComaRec, mn_StartleRes).
- Study 24 has the per-fly data of about 200 lines (StarvationRes, ChillComaRec, StartleRes).
The paper does not publish any per-line values:
- The supplementary material on GSA figshare holds only the supplementary figures and two tables listing the genes in the three most predictive GO terms (Tables S1 and S2).
- The analysis code on GitHub reads the per-sex phenotype files but does not include them. For all trait data, the paper points to the DGRP2 website (http://dgrp2.gnets.ncsu.edu).
All results (prediction accuracies, variance components, predictive GO terms) are given per model, trait, sex or GO term, never per DGRP line, so there is nothing to upload to DGRPool.
The RNA-seq data are in GEO
GSE117850.