Comments from curator
Phenotyping data come from the file
data_all.csv of the authors' Dryad deposit (
doi:10.5061/dryad.f8320d5; downloaded from its Zenodo mirror,
record 4935261, same md5), with the authors' R analysis (Analysis of DGRP Data.Rmd). The file has one row per fly (7,911 flies): lineage (L + RAL number), starvation code (Day), sex, right wing area (WS) and pupal case area (BS) in mm², experimental block, and the natural logs of the two areas used in all the analyses. The supplementary material of the paper (
journal page, also on
PMC) only has figures.
Design (paper): eggs of each DGRP lineage were collected over 12-20 h and transferred in lots of 50 into 7 ml of food, and the larvae were reared at 22 °C. Larvae were removed from the food at timed developmental stages and starved for 0-24 h (
Day 0 flies) or 48-62 h (
Day 2 flies) before pupation, so that each lineage expressed the full range of body and wing sizes. Pupae were kept in individual tubes, and the right wing and the pupal case of each fly were imaged. The lineages were spread over six experimental blocks (incomplete block design). The authors fitted, for each lineage and sex, the scaling relationship between log wing area and log pupal case area (major axis and OLS regressions), pooling all the flies of the lineage, and found that the slopes of these "individual cryptic scaling relationships" vary among lineages and are correlated between the sexes.
Raw dataset 1: one value per fly, 7,906 flies of 87 lines. The wing area (
WingArea_) and the pupal case area (
PupaArea_) of the same fly are on the same row (mm², the WS and BS values exactly as in the file), with one pair of columns per starvation code:
_Day0,
_Day1,
_Day2 and
_DayUnk. The file has four starvation codes: D0 (2,466 flies, Day 0), D2 (2,298, Day 2),
D1 (1,346 flies, blocks 2-4; not described in the paper; their sizes are intermediate between D0 and D2 flies, so presumably an intermediate starvation of about 24-48 h) and
NA (1,801 flies, blocks 4 and 7; starvation level not recorded, a mixture with mean sizes between those of D0 and D2 flies). The authors used all of them in the scaling analyses and the D0 flies alone for the "unstarved" means of Table 3. The flies of the different codes are independent individuals, stacked compactly: row k of a line and sex holds the k-th fly of each code (in file order), and
Block_ gives the block of that fly (obsolete design columns). _DayUnk holds the flies of unknown starvation level. Per code: Day0 2,464 flies, 58 lines; Day1 1,343 flies, 32 lines; Day2 2,298 flies, 70 lines; DayUnk 1,801 flies, 37 lines.
Not uploaded: the natural-log columns (wing, pupa) and four unnamed columns (the areas in µm² and their logs), which are exact transformations of WS and BS; the "unique" fly label (not unique) and the "source" data-sheet label. Dropped: the 5 flies without sex (the authors dropped them too). All the other flies are uploaded, including the line × sex groups with six or fewer flies that the authors excluded from their analyses. L12, L35 and L192 (95 flies) are stored as DGRP_012, DGRP_035 and DGRP_192: RAL-12, RAL-35 and RAL-192 are not in Freeze 2 and were added to DGRPool for this study (not genotyped); the DGRP3 lines RP_0012 and RP_0035 are different lines.
Checks: the file reproduces the paper exactly with the authors' filters (sex f/m; lineage × sex with more than six flies: 81 male and 83 female lineages, the sma test df of 80 and 82; the Fig. 2 legend says 85 and 87): the bivariate means of Table 3 (log mm²: males 0.52 pupa / 0.04 wing for all flies and 0.63 / 0.15 for D0 flies; females 0.61 / 0.23 and 0.74 / 0.36), the denominator df of its Hotelling tests (the male MA row "F2,7556" should read 7656, as in the OLS row), and the correlation between the male and female per-lineage slopes (OLS R² = 0.2263, F(1,78) = 22.82; MA R² = 0.1322, slope 0.95).
The per-lineage slopes and intercepts are derived values shown only in figures (Figs. 2 and 3, Supplementary Figs. S1 and S2) and are not uploaded. To recompute them: per line and sex, regress ln(wing area) on ln(pupal case area) over all the flies of the line (all starvation codes pooled), keeping the line × sex groups with more than six flies, by major axis (R package smatr) or OLS regression. Also not uploaded: the selection data of Frankino et al. 2005 (
Bicyclus anynana) and Wilkinson 1993 (
Cyrtodiopsis dalmanni) reanalysed in this review (same Dryad deposit), which are not DGRP data.