Comments from curator
Phenotyping data comes from the authors' GitHub repository (
promislowlab/zhao2021), cited in the paper's Data Availability Statement, and from
Supplementary Tables S1 and S2 (ACEL-21-e13548-s001.xlsx) in the
Supporting Information section, also available from
PMC. All assays were done on mated females (sex F) of 20 DGRP lines chosen to span the range of lifespan, in two experimental blocks about three months apart (11 and 9 lines, see the obsolete phenotype Block).
Raw dataset 1, climbing activity (one value per fly), from OrganismalPhenotypes/ClimbingActivity20_Raw.txt: rapid iterative negative geotaxis (RING) assay every week during the first six weeks of adult life, on vials of up to 25 females. For each fly, the height climbed 4 s after the last of 10 taps is divided by the available height of the vial (AdjFlyHeight / AdjMaxHeight), which gives a value between 0 and 1 (9,769 measurements). Climb_Week1 to Climb_Week6 hold the flies of each week, stacked: values on the same row come from different weeks and are not linked (the flies were not tracked individually). ClimbVial_Week1 to ClimbVial_Week6 (obsolete) give the vial (the authors' chamber number) of the fly on the same row. Lines 38, 101 and 385 have no data for weeks 5 and 6.
The authors averaged the flies of each vial, then the vials of each line (the vial was their unit of replication), so each vial counts the same whatever its number of flies (1 to 50; the 3 vial numbers with 35 to 50 flies, e.g. DGRP_026 vial 173 at week 3, probably hold two vials recorded under one number; they are grouped as the authors did). Their line means (Table S2 for weeks 3 to 6, the GitHub file Activity_Genotype20.txt for weeks 1 and 2) are kept in the summary dataset as mn_Climb_Week1Auth to mn_Climb_Week6Auth; averaging the vial means of raw dataset 1 reproduces them exactly. The line means computed by DGRPool weigh each fly the same: they differ from the authors' by 0.003 to 0.006 on average depending on the week, and by at most 0.056 (DGRP_355, week 3), with a correlation above 0.997.
Raw dataset 2, lifespan (one row per female), from OrganismalPhenotypes/Lifespan20_Raw.txt, the census of each vial of about 25 females (5 vials per line, 4 for DGRP_208), transferred and scored every other day. Lifespan is the age at death in days (interval deaths, age in hours / 24, as in the authors' script Demography.R). 332 of the 2,355 females were right-censored by the authors (lost or removed alive, 1 to 45 per line): their Lifespan is NA and their age at censoring is kept in the obsolete phenotype CensorAge. LS_Vial is the vial.
The authors' mean and median lifespans (Table S1) are Kaplan-Meier estimates that account for the censored females. They are kept in the summary dataset as mn_LifespanAuth and md_LifespanAuth, and a Kaplan-Meier fit on raw dataset 2 reproduces them exactly. The mean computed by DGRPool from Lifespan ignores the censored females and is lower by up to 4.8 days (DGRP_440: 30 censored out of 87 females).
Raw dataset 3, reproductive output (one row per vial), from OrganismalPhenotypes/ReproductiveOutput_Raw.txt: the lifespan vials left by the females at days 8 and 12 were kept and their adult offspring counted. As defined by the authors (ReproductiveOutput.R), the value is the square root of the number of offspring per female per day (offspring / live females / days in the vial). The line means are identical to Table S2. RO_Vial is the vial, with the same numbering as LS_Vial. The file also has day-4 counts, which the authors did not use; they were not uploaded.
Raw dataset 4, targeted metabolomics (one row per vial cohort), from Metabolomics/TargetedMetabolomicsAug2019.csv and the sample sheet Metabolomics1_Randomized.txt: 87 metabolites measured by LC-MS/MS on pools of 5 females sampled at 4, 10, 24, 45, 69 and 80 days from parallel vial cohorts (1, 2 or 4 cohorts per line, 181 samples). Each row is one cohort, and each phenotype is named metabolite_Dxx, where xx is the age in days. The compound names are in the phenotype descriptions; their HMDB and KEGG identifiers are in Table S7.
The values follow the authors' normalisation (PreprocessingAndStandardization.R): log10 of the total fragment peak area, centred and scaled within each sample over the 87 metabolites. With these values, the 1,044 regressions of Table S4 (metabolite level on mean lifespan, log alpha and beta, days 4 to 45) are reproduced (relative differences below 1e-6). The authors excluded one abnormal sample (DGRP_440, day 4), which is NA here. Only 19 lines were sampled at day 45, 9 at day 69 and 4 at day 80, because the other lines had died out. The raw mass-spectrometry files and Skyline documents are on
Panorama Public.
Summary dataset: per-line values computed by the authors that DGRPool cannot derive from the raw data. From Table S1: mn_LifespanAuth and md_LifespanAuth (see above), the Gompertz-Makeham parameters fitted with WinModest (GM_LogAlpha = ln alpha, GM_Beta and GM_Makeham, per day), and the predicted log mortality rate at 45 and 60 days (GM_LogMort_D45 and GM_LogMort_D60). From Table S2 and the authors' GitHub: the weekly climbing line means mn_Climb_Week1Auth to mn_Climb_Week6Auth (mean of the vial means, see above; weeks 1 and 2 are not in Table S2 and come from Activity_Genotype20.txt), and the rate of decline of climbing activity (ClimbSlope_W3to6, Table S2, the slope of mn_Climb_Week3Auth to mn_Climb_Week6Auth on age in weeks; missing for the 3 lines without data for weeks 5 and 6).
Of note:
- A previous version of this study held the line values of Tables S1 and S2 in raw dataset 1, with sex NA (identical values). Block and Climb_Week3 to Climb_Week6 kept their IDs, but Climb_Week3 to Climb_Week6 now hold one value per fly, so their line means changed slightly (see above); their previous values are now mn_Climb_Week3Auth to mn_Climb_Week6Auth. The other phenotypes moved to the summary dataset under new names, except RepOutput_Day08 and RepOutput_Day12, which are now derived from raw dataset 3 (RepOutput_D08 and RepOutput_D12, same line means).
- The metabolomics sample sheet puts DGRP_038 in block 2, whereas all the other files put it in block 1.
- The GitHub files also give, per line, the age trajectory of each metabolite (AgeTrajectory.txt, used for Table S5) and the Kaplan-Meier survival at the sampling ages (Pheno_20.txt). These were not uploaded.