{
  "id": 13002586,
  "title": "DiverScan: attention-guided exploratory analysis of animal behaviour across diverse comparative experiments",
  "url": "https://urgent.news/2026/10/08/diverscan-attention-guided-exploratory-analysis-of-animal-behaviour",
  "topic": "science",
  "section": "Science",
  "published": "2026-10-08T00:00:00.000Z",
  "source": {
    "name": "bioRxiv",
    "slug": "biorxiv",
    "url": "https://www.biorxiv.org/content/10.64898/2026.10.01.755818v1?rss=1"
  },
  "original_language": "en",
  "account": "High-throughput tracking provides extensive time-series data of animal behavior, yet interpreting the biological insights hidden in the data remains challenging, even for straightforward control-treatment comparisons. Current analyses often fail to identify the multiple, nonlinearly interconnected differences present in the data and struggle to generalize across diverse recording setups. In this study, the authors introduce DiverScan, a deep-learning framework designed for attention-guided exploratory analysis of behavioral time series. DiverScan disentangles condition-dependent behavioral differences by combining a loss function specifically designed for data exploration with an attention-branch architecture. It also visually highlights condition-specific behavioral features and generates explanatory text using a large language model. To validate its effectiveness, DiverScan was tested on both a synthetic benchmark and datasets from seven laboratories encompassing multiple species and both single and multi-animal behaviors. Beyond recovering known phenotypes, the authors demonstrate that DiverScan assists researchers in generating new hypotheses across these diverse datasets. The tool is designed to be user-friendly, available as Jupyter notebooks with a graphical user interface and flexible input formats, enabling seamless integration into existing behavioral-analysis pipelines.",
  "summary": "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…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}