Study

Curated
Manna et al., 2026
Learning is a fundamental source of individuality
[10.7554/elife.111235]

Description

Individual flies of 88 DGRP lines (both sexes) were tested for operant place learning in a Y-maze, avoiding a green arm paired with electric shock, with unshocked controls: task performance, learning score, velocity, handedness and choices. Learning increased within-line individuality.

Comments from curator

Phenotyping data comes from the authors' GitHub repository jaksiclab/GeneticsOfLearningIndividuality (folder "Data Analysis codes"; archived on Software Heritage), which the paper's data availability statement points to (eLife article):
- pL_green/summary_pL_green_90genotypes.csv: one row per fly of the main experiment (green place learning), with the behaviours measured by the tracking software.
- new_learningscore.csv: the learning score of each fly.
- generated_data/grand_*.csv: the authors' per-line statistics of nine behaviours (means and individuality measures).
- pL_green/full_model_pair_compare_sd_90.csv and pL_blue/full_model_blue_pair_compare_sd_with_control.csv: the relative task performance of each line, from the authors' linear mixed models.

Design. 88 DGRP lines (plus the mutants dnc1 and norpAEE5, not uploaded) were reared for two generations at 26°C (12:12 h light:dark), as two biological replicates (progeny of two independent sets of parents, in separate vials). For each line, 64 flies aged 5 days (2 replicates x 2 sexes x 2 treatments x 8 flies) were tested on one day, each in its own Y-maze of a 64-maze arena, mostly at ZT 0-4 (lines tested between December 2022 and February 2024). One arm of each Y-maze was green and two were blue. After 3 min of free exploration (pre-assay, no shock for any fly), the 20-min assay started: conditioned flies were shocked whenever they were in the green arm (operant place learning), control flies were never shocked.

Raw datasets (one row per fly). Conditioned and control flies are different flies, so they are in two datasets: raw_1 for the conditioned flies (names ending in _Cond) and raw_2 for the control flies (_Ctrl), with the same phenotypes:
- TaskPerf: task performance (Tshocked), the fraction of the 20-min assay spent in the green arm (chance 1/3; lower = avoidance), and TaskPerf_Q1 to Q4 for each 5-min quarter.
- LearnScore: learning score (minus the mean Spearman correlation of the time in, and the visits to, the green arm with time over 10 bins of 2 min; -1 to 1).
- Velocity (mm/s, whole assay), Vel_Q1 (first 5 min), Vel_ShockCol and Vel_NeutCol (on the green arm and on the blue arms), Vel_Pre (pre-assay).
- LeftHand and LeftHand_Pre: locomotor handedness, (left turns - right turns) / total turns, during the assay and the pre-assay.
- Choices and CorrChoices: number of decisions (arm entries), and of correct ones (entries into a blue arm).
- GreenBias: fraction of the pre-assay spent in the green arm (colour bias).
- Design columns, obsolete: Replicate, ArenaPos (position of the Y-maze on the arena, 1-64; it sets which arm is green) and ExpDate (assay date).
Values are copied as written in the authors' files. The paper's other per-fly behaviours are derived from these columns: percent correct decisions = CorrChoices / Choices; relative velocity = Vel_ShockCol - Vel_NeutCol; handedness = |LeftHand|; developmental noise = |LeftHand_Pre|; shock response = |Vel_Q1 - Vel_Pre|. Their line means are in the summary dataset.

Excluded flies. We applied the authors' filter, as written in their analysis code (Genetics_of_learning_individuality.qmd). A fly was removed if its velocity was in the lowest 5% of its line x treatment group, or if it never entered the green arm, or if it spent more than 75% of the assay in it (injured flies, according to the authors). 376 of the 5,615 DGRP flies of the file were removed, and one more fly (DGRP_362, control male) whose velocity on the green arm is missing. That leaves 5,238 flies: 2,611 conditioned and 2,627 control. This matches the paper's N (2,611 and 2,628, or 2,627 in Figure 3-figure supplement 1) and Supplementary file 6, cell by cell (except that DGRP_362 fly). The authors' per-line means of all nine behaviours are reproduced exactly from these flies (sexes pooled, max difference 6e-15). The paper's Methods say the filtering removed 3.78% of the wild-type flies, but the code and the legend of Supplementary file 6 (bottom 5% of activity removed in each group) give 6.7%. Supplementary file 6 lists 5,632 DGRP flies measured; the per-fly file has 5,615. Replicates are given as corrected in the authors' code: in the file, 15 lines have flies labelled replicate 3. For DGRP_105 and DGRP_208, replicates 2 and 3 become 1 and 2. For the other 13 lines, the replicate-3 flies are assigned to replicate 1 or 2 from their position on the arena.

Summary dataset (sex NA: both sexes pooled, as in the paper). Per-line values published by the authors in the repository:
- RelTaskPerf: relative task performance, the difference between the least-squares mean Tshocked of the control and of the conditioned flies of the line (positive = learned avoidance). It comes from the authors' mixed model on the 88 DGRP lines (Supplementary file 9), the same values as in their file LSM_frac_dgrpool.tsv. RelTaskPerf_Blue: the same measure in the blue place-learning experiment (the shock is in the blue arm), on 12 DGRP lines (Figure 1-figure supplement 1C, Supplementary file 8; model including dnc1 and norpAEE5).
- mn_PctCorrect, mn_RelVel, mn_AbsHanded, mn_DevNoise, mn_ShockResp (_Cond, _Ctrl): line means of the derived per-fly behaviours listed above.
- Individuality measures, for nine behaviours (TaskPerf, LearnScore, PctCorrect, RelVel, AbsHanded, Velocity, GreenBias, DevNoise, ShockResp):
  - H_: the extent of individuality, the differential entropy (nats) of the distribution of the behaviour among the flies of the line, in conditioned (_Cond) and control (_Ctrl) flies, over the behaviour's theoretical range (Supplementary file 12).
  - Hres_: residual individuality Hresid, the Shannon entropy of the distribution binned into 10 bins over its own range (the shape of the distribution, independent of its variance).
  - dHres_: the change in residual individuality between the conditioned and control flies of the line (ΔHresid Experimental, a Kullback-Leibler divergence).
  - Hell_: the Hellinger distance between the conditioned and control distributions.
DGRPool computes the means, SDs and CVs of the raw phenotypes by line and sex itself. Pooled over sexes, these means equal the authors' values.

Not uploaded:
- the validation experiments (green and blue place and colour learning, multiday learning), which used a single DGRP line (DGRP_362) with the two mutants;
- the per-fly data of the blue place-learning experiment, which are only in an R binary file (Git LFS) and were not read;
- the model fits and simulations, the per-fly "policy change" from the reinforcement-learning model, and the authors' per-replicate statistics;
- 7 learning scores whose fly is not in the per-fly file.
Minor mismatch: the paper gives 82 decisions on average for males, and the uploaded flies average 89.5 (62.4 for females, as in the paper; 63.3 conditioned and 89.3 control, as in the paper).

2 categories

Behaviour Locomotion

98 phenotypes

Choices_Cond Choices_Ctrl CorrChoices_Cond CorrChoices_Ctrl GreenBias_Cond GreenBias_Ctrl H_AbsHanded_Cond H_AbsHanded_Ctrl H_DevNoise_Cond H_DevNoise_Ctrl H_GreenBias_Cond H_GreenBias_Ctrl H_LearnScore_Cond H_LearnScore_Ctrl H_PctCorrect_Cond H_PctCorrect_Ctrl H_RelVel_Cond H_RelVel_Ctrl H_ShockResp_Cond H_ShockResp_Ctrl H_TaskPerf_Cond H_TaskPerf_Ctrl H_Velocity_Cond H_Velocity_Ctrl Hell_AbsHanded Hell_DevNoise Hell_GreenBias Hell_LearnScore Hell_PctCorrect Hell_RelVel Hell_ShockResp Hell_TaskPerf Hell_Velocity Hres_AbsHanded_Cond Hres_AbsHanded_Ctrl Hres_DevNoise_Cond Hres_DevNoise_Ctrl Hres_GreenBias_Cond Hres_GreenBias_Ctrl Hres_LearnScore_Cond Hres_LearnScore_Ctrl Hres_PctCorrect_Cond Hres_PctCorrect_Ctrl Hres_RelVel_Cond Hres_RelVel_Ctrl Hres_ShockResp_Cond Hres_ShockResp_Ctrl Hres_TaskPerf_Cond Hres_TaskPerf_Ctrl Hres_Velocity_Cond Hres_Velocity_Ctrl LearnScore_Cond LearnScore_Ctrl LeftHand_Cond LeftHand_Ctrl LeftHand_Pre_Cond LeftHand_Pre_Ctrl RelTaskPerf RelTaskPerf_Blue TaskPerf_Cond TaskPerf_Ctrl TaskPerf_Q1_Cond TaskPerf_Q1_Ctrl TaskPerf_Q2_Cond TaskPerf_Q2_Ctrl TaskPerf_Q3_Cond TaskPerf_Q3_Ctrl TaskPerf_Q4_Cond TaskPerf_Q4_Ctrl Vel_NeutCol_Cond Vel_NeutCol_Ctrl Vel_Pre_Cond Vel_Pre_Ctrl Vel_Q1_Cond Vel_Q1_Ctrl Vel_ShockCol_Cond Vel_ShockCol_Ctrl Velocity_Cond Velocity_Ctrl dHres_AbsHanded dHres_DevNoise dHres_GreenBias dHres_LearnScore dHres_PctCorrect dHres_RelVel dHres_ShockResp dHres_TaskPerf dHres_Velocity mn_AbsHanded_Cond mn_AbsHanded_Ctrl mn_DevNoise_Cond mn_DevNoise_Ctrl mn_PctCorrect_Cond mn_PctCorrect_Ctrl mn_RelVel_Cond mn_RelVel_Ctrl mn_ShockResp_Cond mn_ShockResp_Ctrl

Datasets

Dataset Phenotypes DGRP lines
Summary dataset 66 phenotypes
H_AbsHanded_Cond H_AbsHanded_Ctrl H_DevNoise_Cond H_DevNoise_Ctrl H_GreenBias_Cond H_GreenBias_Ctrl H_LearnScore_Cond H_LearnScore_Ctrl H_PctCorrect_Cond H_PctCorrect_Ctrl H_RelVel_Cond H_RelVel_Ctrl H_ShockResp_Cond H_ShockResp_Ctrl H_TaskPerf_Cond H_TaskPerf_Ctrl H_Velocity_Cond H_Velocity_Ctrl Hell_AbsHanded Hell_DevNoise Hell_GreenBias Hell_LearnScore Hell_PctCorrect Hell_RelVel Hell_ShockResp Hell_TaskPerf Hell_Velocity Hres_AbsHanded_Cond Hres_AbsHanded_Ctrl Hres_DevNoise_Cond Hres_DevNoise_Ctrl Hres_GreenBias_Cond Hres_GreenBias_Ctrl Hres_LearnScore_Cond Hres_LearnScore_Ctrl Hres_PctCorrect_Cond Hres_PctCorrect_Ctrl Hres_RelVel_Cond Hres_RelVel_Ctrl Hres_ShockResp_Cond Hres_ShockResp_Ctrl Hres_TaskPerf_Cond Hres_TaskPerf_Ctrl Hres_Velocity_Cond Hres_Velocity_Ctrl RelTaskPerf RelTaskPerf_Blue dHres_AbsHanded dHres_DevNoise dHres_GreenBias dHres_LearnScore dHres_PctCorrect dHres_RelVel dHres_ShockResp dHres_TaskPerf dHres_Velocity mn_AbsHanded_Cond mn_AbsHanded_Ctrl mn_DevNoise_Cond mn_DevNoise_Ctrl mn_PctCorrect_Cond mn_PctCorrect_Ctrl mn_RelVel_Cond mn_RelVel_Ctrl mn_ShockResp_Cond mn_ShockResp_Ctrl
88 DGRP lines
DGRP_021
DGRP_026
DGRP_031
DGRP_032
DGRP_059
DGRP_075
DGRP_091
DGRP_093
DGRP_100
DGRP_101
DGRP_105
DGRP_109
DGRP_129
DGRP_153
DGRP_189
DGRP_195
DGRP_208
DGRP_217
DGRP_227
DGRP_228
DGRP_303
DGRP_304
DGRP_309
DGRP_310
DGRP_313
DGRP_317
DGRP_318
DGRP_319
DGRP_321
DGRP_338
DGRP_354
DGRP_362
DGRP_373
DGRP_380
DGRP_382
DGRP_385
DGRP_386
DGRP_390
DGRP_391
DGRP_392
DGRP_395
DGRP_405
DGRP_437
DGRP_440
DGRP_486
DGRP_491
DGRP_502
DGRP_505
DGRP_517
DGRP_535
DGRP_551
DGRP_559
DGRP_563
DGRP_584
DGRP_595
DGRP_627
DGRP_705
DGRP_712
DGRP_716
DGRP_721
DGRP_732
DGRP_737
DGRP_738
DGRP_765
DGRP_776
DGRP_786
DGRP_787
DGRP_790
DGRP_801
DGRP_802
DGRP_804
DGRP_805
DGRP_808
DGRP_818
DGRP_820
DGRP_821
DGRP_822
DGRP_850
DGRP_853
DGRP_855
DGRP_859
DGRP_861
DGRP_879
DGRP_884
DGRP_892
DGRP_897
DGRP_911
DGRP_913
Download
Raw dataset #1 16 phenotypes
Choices_Cond CorrChoices_Cond GreenBias_Cond LearnScore_Cond LeftHand_Cond LeftHand_Pre_Cond TaskPerf_Cond TaskPerf_Q1_Cond TaskPerf_Q2_Cond TaskPerf_Q3_Cond TaskPerf_Q4_Cond Vel_NeutCol_Cond Vel_Pre_Cond Vel_Q1_Cond Vel_ShockCol_Cond Velocity_Cond
88 DGRP lines
DGRP_021
DGRP_026
DGRP_031
DGRP_032
DGRP_059
DGRP_075
DGRP_091
DGRP_093
DGRP_100
DGRP_101
DGRP_105
DGRP_109
DGRP_129
DGRP_153
DGRP_189
DGRP_195
DGRP_208
DGRP_217
DGRP_227
DGRP_228
DGRP_303
DGRP_304
DGRP_309
DGRP_310
DGRP_313
DGRP_317
DGRP_318
DGRP_319
DGRP_321
DGRP_338
DGRP_354
DGRP_362
DGRP_373
DGRP_380
DGRP_382
DGRP_385
DGRP_386
DGRP_390
DGRP_391
DGRP_392
DGRP_395
DGRP_405
DGRP_437
DGRP_440
DGRP_486
DGRP_491
DGRP_502
DGRP_505
DGRP_517
DGRP_535
DGRP_551
DGRP_559
DGRP_563
DGRP_584
DGRP_595
DGRP_627
DGRP_705
DGRP_712
DGRP_716
DGRP_721
DGRP_732
DGRP_737
DGRP_738
DGRP_765
DGRP_776
DGRP_786
DGRP_787
DGRP_790
DGRP_801
DGRP_802
DGRP_804
DGRP_805
DGRP_808
DGRP_818
DGRP_820
DGRP_821
DGRP_822
DGRP_850
DGRP_853
DGRP_855
DGRP_859
DGRP_861
DGRP_879
DGRP_884
DGRP_892
DGRP_897
DGRP_911
DGRP_913
Download
Raw dataset #2 16 phenotypes
Choices_Ctrl CorrChoices_Ctrl GreenBias_Ctrl LearnScore_Ctrl LeftHand_Ctrl LeftHand_Pre_Ctrl TaskPerf_Ctrl TaskPerf_Q1_Ctrl TaskPerf_Q2_Ctrl TaskPerf_Q3_Ctrl TaskPerf_Q4_Ctrl Vel_NeutCol_Ctrl Vel_Pre_Ctrl Vel_Q1_Ctrl Vel_ShockCol_Ctrl Velocity_Ctrl
88 DGRP lines
DGRP_021
DGRP_026
DGRP_031
DGRP_032
DGRP_059
DGRP_075
DGRP_091
DGRP_093
DGRP_100
DGRP_101
DGRP_105
DGRP_109
DGRP_129
DGRP_153
DGRP_189
DGRP_195
DGRP_208
DGRP_217
DGRP_227
DGRP_228
DGRP_303
DGRP_304
DGRP_309
DGRP_310
DGRP_313
DGRP_317
DGRP_318
DGRP_319
DGRP_321
DGRP_338
DGRP_354
DGRP_362
DGRP_373
DGRP_380
DGRP_382
DGRP_385
DGRP_386
DGRP_390
DGRP_391
DGRP_392
DGRP_395
DGRP_405
DGRP_437
DGRP_440
DGRP_486
DGRP_491
DGRP_502
DGRP_505
DGRP_517
DGRP_535
DGRP_551
DGRP_559
DGRP_563
DGRP_584
DGRP_595
DGRP_627
DGRP_705
DGRP_712
DGRP_716
DGRP_721
DGRP_732
DGRP_737
DGRP_738
DGRP_765
DGRP_776
DGRP_786
DGRP_787
DGRP_790
DGRP_801
DGRP_802
DGRP_804
DGRP_805
DGRP_808
DGRP_818
DGRP_820
DGRP_821
DGRP_822
DGRP_850
DGRP_853
DGRP_855
DGRP_859
DGRP_861
DGRP_879
DGRP_884
DGRP_892
DGRP_897
DGRP_911
DGRP_913
Download
Back to studies