Comments from curator
Phenotyping data comes from the authors' Zenodo archive (
10.5281/zenodo.14451399, record 14451400), files
wound_SURVIVAL.csv and
wound_CFU.csv of data.zip. The R analysis of these files is in HR_PLUD.pdf of the same record (section 'Wounds increase both susceptibility to infection and PLUD'). The other files of data.zip (ABAB, Drs_Sk1, RNAi) are AMP mutants and RNAi lines, not DGRP lines.
Design (Fig. 7 and Appendix 7 of the
paper): mated males of 4 DGRP lines (RAL-559, 584, 630 and 818, all with the same
Diptericin A S69 allele), 5-8 days old, reared on glucose-yeast food at 24 °C, were either pricked in the thorax (wound) or not, then injected in the abdomen with 23 nL of
Providencia rettgeri (about 900-2,200 bacteria) or of PBS. Dead flies were collected every 30 min from 12 to 24 h after injection and homogenised one by one in 250 µL of PBS; dilutions were plated as 5-µL drops and the colonies counted. Survival was then followed for up to about 7 days. The experiment was run on 6 injection dates (the authors' random block); PBS controls only on 15/09/2021.
Raw dataset (raw_1), one row per fly (2,765 males; sex M): the columns of a row are filled only for the fly's own condition (
P. rettgeri or PBS, wound or no wound; the other columns are NA).
-
SurvTime_<Pr|PBS>_<NoWound|Wound> (hours): column t of wound_SURVIVAL.csv, the time from injection to death, or to censoring for a fly that did not die.
Dead_...: column Censor (1 = died, 0 = censored). As in the authors' Cox models (Surv(t, Censor)), censored flies keep their censoring time: the end of observation of their block (about 136-187 h, but 40 h for the 17/02/2021 block), 10.45 h for 23 wounded infected RAL-584 flies of 15/09/2021, and about 136 h for 40 unwounded infected flies sampled alive for the set-point load (SPBL; their counts are not in the files). The DGRPool mean of SurvTime therefore mixes death and censoring times; the mean of Dead is the proportion of flies that died.
-
PLUD_Pr_<NoWound|Wound>: log10 pathogen load upon death of the infected flies that died and were plated (389 flies). It is computed from the drop counts of wound_CFU.csv exactly as the authors' script does: CFUfit::loadEstimate(BACTERIA, 1/(4*Dilution), Volume, sample_ID, maxCFU = 50, ignoreAboveMaxCFU = FALSE, logbase = 10). No drop has more than 50 colonies, so this maximum-likelihood estimate is log10(total colonies / sum over drops of Volume/(4 × Dilution)), written at full precision (R's uniroot stops at a tolerance of about 10
-4). It is on the authors' scale (Fig. 7B, 'log_10 PLUD'). Dead flies without drop counts (32) are NA. The authors flagged low PLUD values as outliers with a Rosner test per line and wound treatment (14 flies: 8 unwounded, 6 wounded) and analysed the data with and without them; all values are uploaded. Fly-to-sample matching: 24 sample_IDs (wells of plates 64 and 65) are given to two flies each in wound_SURVIVAL.csv; the counts were assigned by sample_ID, line, wound treatment, injection date and death time, which identifies one fly each (the first one, as in the authors' match()).
-
InjDate,
Replicate,
Inoculum (obsolete design columns): injection date written YYYYMMDD (the authors' block, Day_of_injection), replicate number within date (none for 15/09/2021), and the estimated number of viable bacteria injected per fly on that date (CFU_injected; NA for PBS).
Summary dataset (sex M): the authors' per-line wound effects of Fig. 7C, as printed in HR_PLUD.pdf (d.lHR.lPR).
hr_log_Wound: natural log of the hazard ratio of wounded vs unwounded flies, coefficient Line:Treatment of coxme(Surv(t, Censor) ~ -1 + Line + Line:Treatment + (1|Day_of_injection)) fitted on all flies of the file (
P. rettgeri and PBS pooled, as in the script).
PLUDlogRatio_Wound: log10 ratio of the PLUD of wounded vs unwounded flies, coefficient Line:Treatment of a Gaussian mixed model of the PLUD (weights 1/SE
2, random injection date) without the outliers. The bootstrap 95% CIs are in the PDF and are not uploaded.
Checks: the uploaded data reproduce the authors' outputs exactly: the proportion of dead flies per line, dose and wound treatment (HR_PLUD.pdf p. 1), the numbers of infected flies per line (Fig. 7A legend, e.g. 584: 231/315), the numbers of PLUD estimates per line and treatment (36/52, 74/67, 45/47, 40/28) and per line (Fig. 7B: 141, 88, 68, 92).
The paper is mostly a theoretical model of within-host infection dynamics; the AMP-mutant,
Catalase RNAi and simulation parts do not use DGRP lines. Wounding sharply reduced survival in all lines and increased the PLUD (significant only in RAL-630 when the lines are tested separately), with no significant line x wound interaction.