The paper wouldn't have come together without your bike 🤣

Thanks to the following people/orgs for supporting this work! @carloscaldas1960.bsky.social @caldaslab.bsky.social @cruk-ci.bsky.social @helloeacr.bsky.social @umaryland.edu @churchillcol.bsky.social

Overall, our results suggest characterizing microenvironment composition in the clinic may provide critical prognostic markers that can complement current molecular classification for treatment selection and preventing recurrence.

Paired primary tumors and metastases showed distinct TME compositions depending on their location. Interestingly, naïve B cells were consistently downregulated across all sites. Perhaps this means suppressing the adaptive B-cell response may be a key step for metastatic progression?

We found that tumor–microenvironment interactions significantly shape chemotherapy response, depending on ER subtype. ER– tumors relied more on mature, adaptive immune cells, while ER+ tumors showed a more innate/early-adaptive response pattern.

Microenvironment features also shaped relapse-free survival in context-dependent manners. In particular, certain cell types—like immature perivascular-like cells and plasmacytoid dendritic cells—appeared to influence late recurrence (>10 years), with effects varying by ER subtype.

We identified 7 distinct tumor microenvironment (“TME”) Types that capture immune, stromal, and vascular patterns, normalized for tumor cellularity. The TME Types independently predicted disease-specific survival beyond genomic and intrinsic subtype.

We first benchmarked 15 tumor microenvironment deconvolution methods in 693 breast cancers with matched imaging mass cytometry from the METABRIC cohort. InstaPrism (cousin of BayesPrism) consistently performed best—so we used it for the full 14,837-sample cohort.

We created a meta-cohort of 14,837 breast cancers with clinically annotated RNA-seq data. Using deconvolution, we asked whether the tumor microenvironment could predict outcomes independently of genomic and intrinsic subtype.

The tumor microenvironment—the non-cancerous part of a tumor—plays a major role in patient outcomes and treatment response, yet it’s rarely examined in the clinic.

We asked: can characterizing the microenvironment be clinically useful if done on top of current tumor subtyping?