Spot · whole-slide · atlas

Spatial molecular reasoning from routine histology

HistAgent combines a visual-omics foundation model with a spatial agentic module to support spatial molecular analysis, interactive biological analysis, whole-slide clinical prediction and atlas-scale retrieval.

No sign-in required. Start with the public example or your own image.

Reproduce the analyses with our tutorials
Renal cell carcinoma tissue section used in the HistAgent TLS case study
RCC · H&E specimen
Renal cell carcinoma TLS example

HistAgent localizes and interprets a TLS-like immune niche.

  • 2.23Mpaired H&E–ST spots
  • 936human and mouse slides
  • 32tissue categories
  • 50,000evidence-reasoning traces

Interactive workspaces

Start with tissue.
Follow the evidence.

Analyze a selected tissue location, or retrieve related molecular states from the measured spatial transcriptomics evidence bank.

Reproduce the manuscript analyses

Five notebooks cover spatial biological findings, molecular recovery and ST analyses, clinical prediction and atlas retrieval.

Open tutorials
HistAgent visual-omics foundation model and spatial agentic module
HistAgent model and spatial analysis workflow.View full figure

Method at a glance

A unified framework for spatial molecular analysis

HistAgent couples a visual-omics foundation model with a spatial agentic module to connect local and contextual H&E morphology with question-driven analysis of local tissue states.

  • Local and contextual H&E

    The foundation model jointly encodes a selected tissue location and its surrounding tissue context.

  • Spatial molecular analysis

    HistAgent recovers spatial biological findings and supports standard ST analyses from routine histology.

  • Interactive biological analysis

    The spatial agentic module selects and integrates question-relevant molecular and spatial evidence across multiple turns.

  • Cross-scale applications

    HistAgent extends from spot-level analysis to whole-slide clinical prediction and retrieval from a measured ST evidence bank.

Runnable notebooks

Five tutorials

Each notebook combines explanation, runnable code and outputs. Pretrained models are provided where training would take too long for a tutorial.

  1. Generate and evaluate ranked molecular readouts

    Calculate spot-level gene-ranking recovery and gene-level spatial recovery across five held-out slides.

    HitRate@50 · mAP@50 · PCC
  2. Analyze spatial biological findings

    Compare predicted and measured expression, localize an RCC TLS-like niche and review cross-study finding recovery.

    RCC TLS case study
  3. Run standard spatial transcriptomic analyses

    Run SVG detection, spatial domain identification, deconvolution, differential expression and pathway enrichment.

    Five standard ST analyses
  4. Interpret whole-slide clinical predictions

    Inspect tissue regions associated with slide-level predictions and compare patient-level risk groups.

    WSI · prognosis
  5. Search the spatial transcriptomics atlas

    Explore measured tissue locations and run representative natural-language and H&E image queries.

    Text and image retrieval