← Back to Home

Why Can’t T Cells Enter Tumors? SPACER Looks for Mucin Clues in Tissue Spatial Organization

After analyzing 37 spatial transcriptomics datasets, researchers found links between mucins, T-cell exclusion, and poorer responses to immunotherapy. This interpretable AI provides testable biological clues, but a gap remains before it can inform clinical decisions.

By SURL BioNews

Immune cells may be present near a tumor without being able to penetrate it. For cancer immunotherapy, that distance may be critical. A study led by MD Anderson Cancer Center used the artificial intelligence tool SPACER to analyze gene activity and cell locations within tissues, looking for local signals associated with T-cell accumulation and features accompanying their exclusion.

The study, published in Nature Biomedical Engineering, used 37 spatial transcriptomics datasets. These technologies measure gene expression while preserving information about tissue location; SPACER analyzes neighboring cells together, allowing researchers to trace the local clues underlying the model’s assessments.

According to an explanation released by the institution on October 13, SPACER performed 4 to 21 times as well as existing methods in benchmark tests for identifying biologically meaningful cell recruitment signals. This figure describes comparative results on research tasks and cannot be interpreted as an equivalent improvement in diagnostic accuracy or treatment outcomes.

One set of clues identified by the model involved antigen presentation: expression of genes related to displaying protein fragments for recognition by immune cells accompanied greater T-cell accumulation. Another involved mucins. In melanoma and gastric cancer data, the study found that higher mucin levels were associated with poorer responses to immune checkpoint inhibitors and poorer patient outcomes, supporting further investigation into whether these proteins, which are rich in sugar chains, form a barrier that blocks T cells.

However, spatial associations are not sufficient to establish causality. Developing mucins as therapeutic targets would still require confirmation that intervening against them can change T-cell entry into tumors; using them as biomarkers would also require validation in independent patient data. The information released to date does not report that SPACER has delivered clinical benefits.

Data access is also part of assessing whether the results can be reproduced. Zenodo hosts a dataset bearing the same title as the study, listed as published on July 31, version v1, with Yunhe Liu of MD Anderson named as the data collector. The record is publicly viewable, but the files are restricted; external researchers cannot obtain the data from that page alone, making it difficult to directly rerun the full analysis using that resource. Moving from hypotheses drawn from tissue images to evidence of improved patient treatment still requires follow-up experiments and clinical validation.

References

  1. MD Anderson Cancer Center via Newswise
  2. The University of Texas MD Anderson Cancer Center
  3. Zenodo