Comments from curator
This study was first submitted to DGRPool as the bioRxiv preprint (doi
10.1101/533711); it was curated from the published version (
Science Advances 2020, 6(8):eaay3047,
PMC7034997).
Phenotyping data come from the authors' Dryad dataset
doi:10.5061/dryad.3xsj3txcn (files taken from its identical open copy on
Zenodo record 4976567, MD5 checked), files
20.04.17.panel.longswitch.csv and
20.04.17.panel.4day.csv (the DGRP panel experiment), and from
Tables S8 and S11 of the
Supplementary Materials (aay3047_SM.pdf).
Assay: mated females of 11 DGRP lines (DGRP_105, 136, 195, 217, 239, 335, 362, 441, 705, 707 and 853) were kept in demography cages of about 125 females (100 to 150) at 25 °C, 12:12 h light/dark, 50-60% relative humidity. The rich food contained 8% autolysed yeast, 13% sugar, 6% cornmeal and 1% agar; the restricted food (DR) was identical but with 2% yeast. Females entered the cages at 2 days of age and stayed on rich food until the first census (age 3 or 4 days). Each cage then received one of four regimes:
Rich (continuous rich food; 10 cages per line, 8 for DGRP_195),
DR (continuous restricted food; 6 cages),
LongSw (long switch: DR, then rich food for the rest of life from the age at which 45-65% of the continuous-rich females were still alive, i.e. from 17-19 days for DGRP_707 to 43-45 days for DGRP_441; 6 cages) and
Sw4d (4-day switch: DR and rich food alternating every 4 days, 3 cages starting on DR and 3 starting on rich food; 6 cages). A census was taken every other day: dead females were counted and removed and the food was renewed. Females found alive but stuck to the cage or food, and escaped ones, were right-censored. In total 37,897 females were followed, as stated in the paper.
Processing: the Dryad files are in the counting-process format of the authors' time-dependent Cox models: one row per female and per 2-day census interval during which she was at risk (columns Genotype, Cage, Age, Treatment, Dead, Diet), with no fly identifier. The continuous rich and DR cages are the same flies in both files (identical rows) and were used once. Within each cage, the females found dead at a census are the sum of Dead at that Age, and the females censored at a census are those at risk at that census minus those at risk at the next census minus the deaths. This rebuilds every female: 37,897 in total, with 13,375 Rich, 8,170 DR, 8,157 LongSw and 8,195 Sw4d (29,702 for Fig. 3 and 29,740 for Fig. S1, exactly as in the legends). Re-expanding the uploaded file gives back the at-risk and death counts of the source files for all 306 cages and census ages.
Raw dataset 1 (one row per female, sex F): the four regimes are independent cages, so they are stacked compactly: row k of a line holds the k-th female of each regime (ordered by cage, then age).
-
Lifespan_Rich,
Lifespan_DR,
Lifespan_LongSw,
Lifespan_Sw4d: age in days (Age column of the source files) at the census at which the female was found dead. Lifespan_LongSw includes the females that died on DR before the switch.
-
CensorAge_* (obsolete): age at the last census of the 2,332 right-censored females (6.6% of the deaths; the Methods give ~10.5% of deaths, apparently over all experiments). Their Lifespan is NA, so the line means computed by DGRPool are means of the observed ages at death and ignore the censored females, whereas the authors' models treat them as right-censored.
-
Cage_* (obsolete): the authors' cage number (numbered within each line; the authors fitted cage as a random effect).
-
SwitchAge_LongSw (obsolete): age at the census at which the female's cage was moved from DR to rich food.
FirstDiet_Sw4d (obsolete): yeast content (2 or 8%) of the food of the first 4-day period of the female's cage. Both are read from the Diet column of the source files, so that the authors' interval models can be refitted.
Summary dataset (sex F): the per-line log hazard ratios of the authors' within-genotype interval-based Cox models (continuous rich food as reference, cage as random effect), as printed in the column "estimate" of Tables S8 and S11 (summary type hazard ratio):
-
hr_log_DR_ExpLS (Table S8, "DR"): mortality on DR vs. continuous rich food, in the long-switch experiment.
-
hr_log_LongSw (Table S8, "long switch"): mortality of long-switch females after their return to rich food vs. continuous rich food (the mortality cost of returning to a rich diet after DR; positive = higher mortality).
-
hr_log_DR_ExpSw4d (Table S11, "DR"): mortality on DR vs. continuous rich food, in the 4-day switch experiment.
-
hr_log_Sw4d (Table S11, "4-day switch"): mortality of 4-day switch females during their rich-food periods vs. continuous rich food.
-
hr_log_Sw4dxDR (Table S11, "4-day switch * DR"): interaction term; the log hazard ratio of 4-day switch females during their DR periods vs. continuous rich food is hr_log_DR_ExpSw4d + hr_log_Sw4d + hr_log_Sw4dxDR.
In Tables S8 and S11 the columns headed "exp" and "s.e." are swapped (exp(estimate) is printed under "s.e."); the hazard ratio is exp of the uploaded value. Standard errors and p-values were not uploaded. Refitting these models on the source files (gamma frailty on cage instead of the authors' mixed model) gives estimates close to the tables (48 of 52 comparable values within 0.05, median difference 0.01, max 0.21 for hr_log_LongSw of DGRP_217), which confirms that the deposited files are the data analysed in the paper. The authors' panel-wide models (Tables S7, S9 and S10, DGRP_195 as reference) and the refits restricted to the four lines that starved on DR (Tables S12 and S13) were not uploaded.
Not uploaded:
- The experiments done on DGRP_195 only (Fig. 2, Figs. S4-S7, Tables S1-S6 and S18-S21; the other files of the Dryad dataset): the same regimes at high sample size plus 2-day switch, short reverse switch and asymmetric switches, antibiotic treatment, water supplementation, females housed singly, and males.
- Fecundity (QuantiFly relative egg counts per cage, Figs. S2 and S3, Tables S14-S17): these data are not in the Dryad dataset.
Of note:
- The 8 cages of DGRP_853 numbered 21 to 28 have no census at age 37 (one 4-day interval from age 35 to 39).
- The long-switch age differs by up to 3 days between cages of a line because the cohorts eclosed on different days and all cages of a line were switched on the same day.
