Back

DiverScan: attention-guided exploratory analysis of animal behaviour across diverse comparative experiments

Sukegawa, M.; Azechi, H.; Butsugan, T.; Freund, N.; Hakataya, S.; Iwata, T.; Katsu, N.; Kimura, K. D.; Koda, H.; Koto, A.; Makiguchi, Y.; Matsui, H.; Miyoshi, K.; Mundorf, A.; Nishimori, H.; Okanoya, K.; Shiraishi, M.; Takahashi, S.; Tanaka, R.; Toya, G.; Yamazaki, C.; Yamazaki, S. J.; Maekawa, T.

2026-10-08 animal behavior and cognition
10.64898/2026.10.01.755818 bioRxiv
Show abstract

High-throughput tracking yields rich time-series data of animal behaviour, yet extracting biological insight remains a bottleneck, even for simple control--treatment comparisons. Existing analyses struggle to pinpoint the multiple, nonlinearly intertwined differences hidden in the data and fail to generalise across diverse recording set-ups. Here we present DiverScan, a deep-learning framework for attention-guided exploratory analysis of behavioural time series. DiverScan disentangles condition-dependent behavioural differences by pairing a loss function tailored to data exploration with an attention-branch architecture, and highlights condition-specific behavioural features visually and with explanatory text generated by a large language model. DiverScan is validated on a synthetic benchmark and on datasets from seven laboratories spanning multiple species and both single- and multi-animal behaviour. Beyond recovering known phenotypes, we demonstrate that DiverScan helps researchers generate new hypotheses across these diverse datasets. Available as Jupyter notebooks with a graphical user interface and flexible input formats, DiverScan can integrate into existing behavioural-analysis pipelines.

Matching journals

The top 8 journals account for 50% of the predicted probability mass.

1
Nature Communications
○ 5641 papers in training set
Top 20%
9.4%
Similar papers in this journal
2
Cell Reports Methods
○ 165 papers in training set
Top 0.2%
6.4%
Similar papers in this journal
3
Nature Machine Intelligence
○ 70 papers in training set
Top 0.3%
6.4%
Similar papers in this journal
4
Proceedings of the National Academy of Sciences
● 2444 papers in training set
Top 7%
6.4%
Similar papers in this journal
5
eLife
● 5828 papers in training set
Top 19%
6.4%
Similar papers in this journal
6
PLOS Computational Biology
● 1863 papers in training set
Top 6%
6.0%
7
Scientific Reports
○ 3612 papers in training set
Top 17%
5.3%
8
Neuron
○ 337 papers in training set
Top 2%
5.2%
50% of probability mass above
9
Nature Neuroscience
○ 252 papers in training set
Top 2%
5.2%
10
Nature Methods
○ 385 papers in training set
Top 2%
4.6%
11
PLOS Biology
● 486 papers in training set
Top 1%
4.1%
12
Nature
○ 645 papers in training set
Top 4%
3.9%
13
Science Advances
● 1243 papers in training set
Top 10%
3.4%
14
Methods in Ecology and Evolution
○ 176 papers in training set
Top 0.8%
3.0%
15
iScience
○ 1154 papers in training set
Top 14%
2.0%
16
Ecology Letters
○ 135 papers in training set
Top 1%
1.8%
17
PLOS ONE
● 5266 papers in training set
Top 48%
1.7%
18
Science
● 477 papers in training set
Top 6%
1.6%
19
Communications Biology
○ 993 papers in training set
Top 18%
1.4%
20
Cell Reports
○ 1498 papers in training set
Top 26%
0.9%
21
BMC Bioinformatics
○ 457 papers in training set
Top 5%
0.9%
22
Genome Biology
○ 637 papers in training set
Top 9%
0.8%
23
GigaScience
◐ 212 papers in training set
Top 5%
0.8%
24
Journal of Animal Ecology
○ 75 papers in training set
Top 2%
0.8%
25
Nucleic Acids Research
◐ 1281 papers in training set
Top 16%
0.6%
26
Genome Research
● 468 papers in training set
Top 8%
0.6%
27
Cell Systems
○ 201 papers in training set
Top 6%
0.6%

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.