Comments from curator
Phenotyping data comes from the authors'
Zenodo deposit (
10.5281/zenodo.32143, data.zip, file
DGRP_data_Odor_Impulse.mat): the binary walking/stationary time series (20 frames per second, 200 s) of every fly of the 98 DGRP lines, with each fly's odor onset and offset frames. The per-line means were checked against
S1 Table of the
paper (Supporting information, pcbi.1004577.s009.xlsx: RAL number and mean pre-odor spontaneous walking frequency of each line).
Design: adult females (2-5 days old, raised at 25°C on a 12:12 h light:dark cycle, starved for 4-6 h) were tracked individually at 20 fps in 50 x 10 mm arenas under far-red light. After 1 min of acclimation, each fly received 60 s of air flow, 30 s of 10% acetic acid throughout the arena, then air again (the recording lasts 200 s). Walking was scored with a hysteresis threshold on speed (walking starts above 1 mm/s and ends below 0.5 mm/s). Between 131 and 242 flies were measured per line (20,223 flies in total); the 98 lines were screened over about one year, in randomly chosen groups of about 20 lines.
The raw dataset has
one row per fly (sex F) and three phenotypes, each the fraction of frames during which the fly was walking (0 to 1):
-
WalkFreq_PreOdor: basal walking frequency before the odor, over frames 1-1150 (first 57.5 s). This is the authors' measure: the line means computed from it reproduce S1 Table exactly (98/98 lines, maximum difference 8e-15). The window of S1 Table is not stated in the paper; it was found by testing all windows, and only frames 1-1150 reproduce it.
-
WalkFreq_Odor: walking frequency during the 30 s odor pulse (from the fly's odor onset frame to its offset frame).
-
WalkFreq_PostOdor: walking frequency from the end of the odor pulse to the end of the recording (about 110 s). It includes the post-odor decay of locomotion as well as the reduced basal locomotion that follows, so it is not the paper's post-odor basal frequency (Fig 2C).
WalkFreq_Odor and WalkFreq_PostOdor were computed by DGRPool from the authors' time series: the paper shows them only as time courses (Fig 2A), not as per-line values. The odor onset and offset frames of each fly are kept as obsolete columns (OdorOnFrame, OdorOffFrame). Since the raw dataset reproduces S1 Table, no summary dataset was added.
Of note:- The odor onset is at about frame 1200 (60 s) for most flies, but at frames 1303-1304 for 71 flies and at frames 2367-2371 (about 118 s) for 71 other flies, in 20 lines (5-15 flies per line); the walking response of these flies follows their own onset frame. WalkFreq_Odor and WalkFreq_PostOdor use each fly's own frames (for the flies with a late onset, the post-odor window is only about 52 s). WalkFreq_PreOdor (frames 1-1150) is before the odor for every fly. The authors' line time courses average the flies frame by frame, without this realignment.
- RAL208 was screened four times to check reproducibility, but only one run (203 flies) is in the deposit.
- The data folder names also give a Bloomington number. Two do not match DGRPool: 78183 for RAL371 (a typo for 28183) and 28222 for RAL737 (now Bloomington 83729). The lines were identified by their RAL number, which agrees with S1 Table.
- The S7 Fig legend gives 225 flies for RAL371 and RAL642; the deposit has 207 flies for each.
Not uploaded: the duration of the post-odor locomotor decay (Fig 2B) and the ratio of post-odor to pre-odor basal locomotion (Fig 2C), which are shown only in figures and whose computation is not described; the odor aversion test that followed (median odor aversion, Fig 2F-G), which is not in the deposit; and the Canton-S recordings (not a DGRP line). The paper has no GWAS: the DGRP lines were used to constrain neural network models of locomotion.