MD Anderson's TmS Biomarker Outperforms Standard Tools for Triple-Negative Breast Cancer

Triple-negative breast cancer - which lacks estrogen receptor, progesterone receptor, and HER2 amplification - is the subtype with the fewest targeted treatment options and one of the most variable outcomes. Some patients respond dramatically to chemotherapy; others progress on the same regimens. Identifying upfront which patients will respond has proven difficult, in part because the computational tools used to classify tumors have systematically overlooked a key variable.

Researchers at MD Anderson Cancer Center identified that gap and built a new biomarker designed to fill it. The tool, called TmS - for tumor-specific total mRNA expression - accounts for gene expression changes within tumor cells relative to the surrounding microenvironment, a dimension that existing deconvolution approaches largely ignore. Results from 575 patients across ethnically diverse cohorts, published February 17 in Cell Reports Medicine, suggest TmS predicts chemotherapy response more accurately than current standard methods.

The gap in current deconvolution methods

Bulk RNA sequencing of tumor samples captures gene expression from a mixture of cell types: cancer cells, immune infiltrates, stromal cells, blood vessel cells. Deconvolution methods try to computationally separate these contributions and infer the cell composition of the tumor. More than 40 such methods exist, and MD Anderson's group recently published a comprehensive guide cataloguing 43 of them.

The limitation those methods share is that they measure the proportions of different cell types without accounting for how gene expression within the tumor cells themselves changes in response to the surrounding microenvironment. Those context-dependent expression changes can be diagnostically significant - they reflect how a tumor is actually behaving in its specific biological environment, not just what genes it expresses in isolation.

Normal cells express mRNA in amounts proportional to chromosome number. Cancer cells do not - they typically carry abnormal numbers of chromosomes (a condition called aneuploidy), and their gene expression deviates accordingly. TmS incorporates that deviation explicitly, using the ratio of tumor cells to non-tumor cells and adjusting for chromosome-level expression differences to generate a cancer-specific signal.

Performance in 575 patients

The research team, led by Wenyi Wang, PhD, professor of Bioinformatics and Computational Biology at MD Anderson, applied TmS to a dataset of 575 TNBC patients drawn from ethnically diverse cohorts. The biomarker sorted patients into two groups: high-TmS (associated with favorable prognosis) and low-TmS (associated with poor prognosis).

Head-to-head against existing deconvolution methods on chemotherapy response prediction, TmS outperformed the current tools. The researchers do not specify exact performance metrics in the press release; full statistical details are available in the Cell Reports Medicine paper.

"Deconvolution strategies are not one size fits all," Wang said. "We're focused on making these methods more accessible to researchers without extensive computational backgrounds, with the goal of translating these powerful analytical approaches into practical tools that the broader cancer research community can readily apply to advance precision medicine."

Population-level differences revealed

A notable secondary finding concerns biological differences between Asian and Western (European-ancestry) patients in the high-TmS group. The biomarker identified distinct tumor microenvironment profiles between these populations - differences in how non-cancer cells surrounding the tumor are organized and active. The authors suggest those distinctions could allow clinicians to better match specific additional therapies - beyond standard chemotherapy - to each population group.

This finding is preliminary but clinically relevant given that TNBC rates and outcomes vary by ancestry, and most breast cancer biomarker research has been conducted predominantly in European-ancestry populations. Tools that explicitly address cross-population variation are needed as precision oncology expands globally.

Validation still required before clinical use

The current work is a research-stage demonstration. The TmS biomarker has not been validated in prospective clinical trials or compared against standard-of-care treatment selection in a controlled setting. Retrospective performance on existing datasets, while encouraging, does not guarantee that deploying the tool prospectively will change outcomes for patients. The authors explicitly note that further validation is needed before clinical implementation.

The study was supported by the National Cancer Institute, the Department of Defense, the Cancer Prevention and Research Institute of Texas, the American Cancer Society, and Lyda Hill Philanthropies.