TRACE: Tumor Reactive T Cell Classifier Scores Across Solid Tumors

A tumor reactivity scoring algorithm (TRACE) with open model weights, released in April 2026. The associated data likely contains TRACE scores applied to hundreds of patient samples from tumor atlases spanning lung, colorectal, and pancreatic cancer. The algorithm was developed by David Monteiro using aggregated scRNA-seq data from multiple publications to predict tumor-reactive CD8+ T cells.

Use Cases

• Characterizing the fraction of tumor-reactive T cells (TRTs) in tumor-infiltrating lymphocytes based on the TRACE algorithm.
• Establishing correlations between TRT frequency and clinical response to immunotherapies as mentioned in the conclusion.
• Identifying tumor-antigen experienced T cells in tissue or blood scRNA-seq datasets using the provided model weights.
• Benchmarking new TRT prediction methods against TRACE's reported performance metrics (Matthews correlation coefficient of 0.84, F1-score of 0.85).

Strengths

• Algorithm performance is quantified with specific metrics: a mean Matthews correlation coefficient of 0.84 and F1-score of 0.85.
• Applied to hundreds of patient samples from multiple tumor atlases, suggesting a broad validation scope.
• Released under a permissive CC-BY-4.0 license with open model weights, facilitating reuse and application.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for specific analyses.
• The primary file format is a PDF (2.1 MB), which may not be the most accessible format for direct computational analysis.

Provenance

Source

figshare

Collection Method

Aggregated dataset of validated tumor reactive clonotypes and associated scRNA-seq data from multiple publications.

Time Range

Publication date suggests data is current as of April 2026.

Freshness

Last updated 2026-04-16 05:16:31; freshness should be verified.

Geography

Samples likely originate from the tumor atlases referenced, but specific geographic coverage is not stated.