Comments from curator
Phenotyping data comes from the authors'
Zenodo deposit cited in the data availability statement (
doi:10.5281/zenodo.21066881, file zonedo08012026.zip):
-
clock_gene_expression_merged.csv (folder background_simulations): relative expression of ten core clock genes for 120 DGRP lines, used for the raw dataset. It is the full-precision copy of
data/expression_normalized.tsv (same lines, genes and missing values; values rounded to 9 decimals there).
-
data/gwas_input.zip: the authors' GWAS input of each gene (natural log of the same values; not uploaded, see below).
Supplementary Tables S2 (correlations among genes) and S3 (eQTL list) of the
Supplementary information were used to check the data.
Design: 120 DGRP lines kept in the authors' laboratory (18 °C, 12:12 h light:dark, cornmeal-yeast-molasses food) were moved to 24 °C, 12:12 h light:dark, and their progeny developed at 24 °C. About 50 mixed-sex adults aged 3-5 days were collected at Zeitgeber time 2 (ZT2, 2 h after lights-on) and frozen. Heads were separated by vortexing and RNA was extracted from the heads. The expression of ten core clock genes (
Clk,
cry,
cwo,
cyc,
dbt,
Pdp1ε,
per,
sgg,
tim,
vri) and of the reference gene
RpL32 was measured by qPCR (SYBR Green, QuantStudio 3). Quantities were read from standard curves made from a pool of cDNA of 20 samples, and the quantity of each gene was divided by that of
RpL32 (relative expression, unitless).
One biological replicate (one pool) was measured per line, at a single time point.
raw_1 has one row per DGRP line (its single pool), with sex NA (mixed-sex pools whose sex ratio was not recorded), and one phenotype per gene, named <gene>_HeadExpr_ZT2 (
Pdp1e_HeadExpr_ZT2 is
Pdp1ε, written "pdp" in the deposit). The values are copied unchanged, so the line means computed by DGRPool are the authors' values. The authors removed outliers gene by gene (the method is not stated); these values, like the missing measurements, are NA. Each gene has 105 to 117 lines (the paper says 100-117), and 83 lines have all ten genes. All 120 lines are DGRP2 lines.
Log transformation: the uploaded values are not log-transformed (as in Fig. 1). The authors' GWAS (DGRP2 web tool, one gene at a time) used their natural logarithm: the files of gwas_input.zip equal ln(value) for the same lines (exactly for six genes, to within 5 × 10
-4 for
Clk,
cyc,
per and
vri), and the effects in Table S3 fit the genotype means on the log scale. The log values are not uploaded, since they are a transform: take the natural log of the uploaded values to reproduce the GWAS input.
The authors also mapped the first two principal components of the standardised expression of the 83 lines with complete data (PC1, 43.3% of the variance:
dbt,
Pdp1ε,
sgg,
cwo,
vri; PC2, 16.5%:
per,
cry; Tables S7 and S8). The per-line PC scores were not published and are not uploaded. They can be recomputed from the uploaded values (R prcomp, centred and scaled, untransformed values, complete lines only).
Checks: the Pearson correlations of the uploaded values (pairwise-complete lines) reproduce Table S2 for all 45 gene pairs, to the 3 decimals given (96 to 115 lines per pair, as stated). The highest/lowest line ratios reproduce the fold changes of Fig. 1 (from 8 for
cry to 86 for
Pdp1ε), and the PCA reproduces the variance explained and the loadings given in the text.
Of note, the text has two errors:- The 86-fold range of
Pdp1ε is between lines 732 and 774 (the text says 744, which is not in the data).
- The text and abstract say that
cyc varied least (11.3-fold, lines 142 and 392). In the data and in Fig. 1,
cyc varies 14-fold (DGRP_357 to DGRP_352); lines 142 and 392 are the extremes of
per (11-fold), and the smallest range is that of
cry (8-fold).
The authors caution that the sex ratio of the pools was not controlled. In the sex-specific DGRP transcriptome of Huang et al. (2015),
Clk,
cyc and
cwo have a significant line × sex interaction (Fig. S7), so these three genes should be interpreted with more caution.
Not uploaded: the raw Ct values of data/expression_matrix.tsv (123 rows without line names, which cannot be matched to the lines), the GWAS results, and the
Mad RNAi knockdown experiment (PDF-GAL4 > UAS-
Mad RNAi and the two parental controls: qPCR and free-running period; not DGRP lines). A preprint of this work is on bioRxiv (
doi:10.1101/2025.05.18.654706).