Comments from curator
Phenotyping data comes from the authors'
companion data page (de Bivort lab), file
vofv_AllYMazeData.mat (MATLAB), which holds the left/right turn data of every fly of the study:
-
expData_DGRPScreen (21,505 flies, Fig. 1): the DGRP screen, used for the raw_1 dataset.
-
expData_VarHerit (1,326 flies, Fig. 2): reciprocal crosses between the high-variability lines 45 and 105 and between the low-variability lines 535 and 796, with their parental lines re-measured at the same time; the parental lines are used for the raw_2 dataset.
- expData_TenaMutants and expData_TenaTimeCourse (Figs. 2-3:
Ten-a null allele, deficiency and RNAi): not DGRP lines, not uploaded.
The supplementary information (Tables S1 and S2: ANOVA and GWAS results) has no per-line values.
Assay: flies reared at 25 °C (12/12 h light/dark) were lightly anesthetized and placed individually in the Y-mazes of a 120-maze array; after a 20 min recovery, each left/right turn was recorded for 2 h. Flies were randomized across blocks, lines, arrays and maze positions, with at least three lines per array. Only females were tested. Each row of the raw datasets is one fly.
Phenotypes (per fly):
-
TurnBias: fraction of the fly's turns that were to the right, as provided in the file (1 − the mean of the stored turn-direction vector, whose left/right coding is flipped by the authors' parser flyY120.m). The main text defines it as the fraction of right turns, the SI text as the fraction of left turns; the trait is symmetric, so this does not change the variability analyses.
-
NumTurns: number of turns in the 2 h (an estimate of locomotor activity). The file only has flies with at least 50 turns (the paper says "more than 50"; the authors' code keeps 50 or more).
-
Switchiness and
TurnRegularity: not stored in the file. They were computed by the curator from each fly's turn-direction and turn-time vectors with the authors' formulas (flyY120DataWorkup.m on the companion page): switchiness = (N
LR + N
RL)/(2N
RN
L/N), the number of alternations relative to a random sequence (1 = random); regularity of turn timing = MAD(inter-turn intervals, s)/(7,200/N), low for flies that turn in a few dense bursts.
-
ExpDate,
ExpTime,
Rig,
Tray and
Maze (obsolete): date (the block of the authors' ANOVA), start time, imaging rig ("box"), tray (maze array) and maze position of the fly.
The paper's trait is the variability of each line, not its mean. The mean turn bias does not differ between lines (every line is unbiased on average), whereas the intragenotypic variability of turn bias, measured by the median absolute deviation (MAD = median of |x − median(x)| over the flies of a line), differs strongly between lines and was the phenotype of the authors' GWAS (36 variants in 22 genes at P < 5 × 10
-6, including
Ten-a). DGRPool computes the per-line mean, median, SD and CV from the raw data, not the MAD, and the authors did not publish per-line MADs. To reproduce the GWAS phenotype, compute the MAD of TurnBias per line over the 159 lines described below.
Lines: raw_1 has 19,348 flies of 199 DGRP lines. The paper analyses 159 lines: the genotyped (Freeze 2) lines with at least 75 flies (109.6 flies per line on average; the paper says 110). The other 40 lines are in the authors' file but not in the paper: 7 lines with at least 75 flies that are not genotyped, DGRP_147 (110), DGRP_221 (118), DGRP_272 (103), DGRP_378 (111), DGRP_387 (103), DGRP_476 (105), DGRP_609 (116), and 33 lines with fewer than 75 flies: DGRP_041 (37), DGRP_055 (1), DGRP_128 (43), DGRP_142 (32), DGRP_152 (12), DGRP_155 (10), DGRP_235 (69), DGRP_301 (13), DGRP_303 (27), DGRP_317 (68), DGRP_332 (9), DGRP_335 (59), DGRP_338 (62), DGRP_360 (57), DGRP_371 (68), DGRP_382 (34), DGRP_384 (14), DGRP_398 (42), DGRP_427 (26), DGRP_487 (28), DGRP_492 (67), DGRP_530 (52), DGRP_531 (3), DGRP_554 (32), DGRP_563 (42), DGRP_595 (40), DGRP_730 (26), DGRP_785 (5), DGRP_788 (9), DGRP_799 (38), DGRP_862 (49), DGRP_884 (52), DGRP_906 (29).
Eleven of these lines are not in DGRP Freeze 2 and were added to DGRPool for this study (not genotyped; 534 flies): DGRP_128 (43), DGRP_147 (110), DGRP_152 (12), DGRP_155 (10), DGRP_221 (118), DGRP_384 (14), DGRP_487 (28), DGRP_609 (116), DGRP_785 (5), DGRP_862 (49), DGRP_906 (29). The authors' labels are RAL numbers, so they are stored as DGRP_ lines; the DGRP3 lines (RP_) with the same numbers are different lines.
Not uploaded: 2,076 Canton-S control flies; 32 flies labelled "799A" (line 799 itself has 38 flies, uploaded as DGRP_799); and 49 flies with 5 labels that are not DGRP lines: 225 (27 flies), 536 (19), 78, 329 and 829 (1 each). Labels 78, 225, 536 and 829 are probably typos (they are not in the experiment's file name, which lists 378, 228, 563 and 892); label 329 has a single fly.
raw_2 has 365 flies of the parental lines DGRP_045 (124) and DGRP_105 (51) (high variability), DGRP_535 (100) and DGRP_796 (90) (low variability), re-measured in September 2012 for the heritability crosses of Fig. 2 (phenotypes with the suffix _Par, defined as in raw_1). The F1 progenies of the reciprocal crosses between two DGRP lines (45 × 105 and 535 × 796, 752 flies) and the Canton-S controls are not uploaded. Of note, the Fig. 2 legend gives other parental sample sizes (105 × 105: 235; 45 × 45: 315; 535 × 535: 197; 796 × 796: 265), which match neither this file nor this file plus the screen flies of the same lines; the F1 sample sizes of the legend match the file.
Checks against the paper (on the 159 lines): Levene's test of the turn-bias variance F = 7.701 (paper 7.701), Brown–Forsythe F = 7.565 (paper 7.567); 29 dates, 6 rigs and 12 trays, as the degrees of freedom of the paper's ANOVA (28, 5 and 11); r
2 between the line means and MADs of turn bias = 0.0004 (paper 0.0004); the
Ten-a SNP X:12153076 splits the lines into 12 minor-allele and 141 major-allele lines, as in the GWAS table; the correlations between the line MADs of the four traits are close to Fig. S3 (MAD of turn bias vs MAD of switchiness: r
2 = 0.11, paper 0.13). The analysed flies made 427 turns on average (paper 413).