
1/ Thrilled to share our new paper, out today in @Nature: "Non-invasive profiling of the tumour microenvironment with spatial ecotypes". Paper (open access): nature.com/articles/s41586-0…

Computational biology, cancer genomics, & stem cell bioinformatics lab @Stanford @czbiohub

1/ Thrilled to share our new paper, out today in @Nature: "Non-invasive profiling of the tumour microenvironment with spatial ecotypes". Paper (open access): nature.com/articles/s41586-0…

An honor to discuss our recent @Nature paper with @EricTopol on his Ground Truths podcast! See his fantastic thread below and a link to our conversation here: erictopol.substack.com/p/a-b…
For treatment and outcomes for cancer, the tumor microenvironment (TME) is pivotal, but we've never had a non-invasive way to assess it. Until now, a blood test breakthrough! (<-a term I use sparingly) Pictured below, schematically, the TME surrounds the cancer cells and consists of immune cells, blood vessel (endothelial) cells, fibroblasts and matrix

16/ GitHub: github.com/digitalcytometry/… Web interface: spatialecotyper.stanford.edu

15/ We are grateful to our funding sources including @theNCI, @TheVFoundation, @CancerRF, @MelanomaReAlli, and @Ludwig_Cancer. We deeply thank the patients and their families whose generous contribution of samples made this research possible.

14/ This work is the result of an incredible collaborative effort across teams at @StanfordMed, @MayoClinic, @washumedicine, and @YaleMed.

13/ This opens the door to deciphering TME dynamics in real time and serial monitoring of the TME in the clinic, with the potential to guide new paradigms for more precise, personalized cancer care.

12/ We then measured liquid spatial ecotypes in nearly 100 patients with melanoma treated with immunotherapy, and showed that pretreatment liquid SE levels were strongly predictive of both immunotherapy response and survival.

11/ We showed that spatial ecotype levels derived from plasma cell-free DNA methylomes correlate well with ground-truth tumor tissue-derived spatial ecotypes, and show specificity against peripheral blood mononuclear cell (PBMC) levels.

10/ Once discovered in spatial data, SEs can be recovered not only from non-spatial modalities (including bulk and scRNA-seq), but non-invasively as well using Liquid EcoTyper, an interpretable AI algorithm that we developed and benchmarked in this study.

9/ Indeed, spatial ecotypes outperformed clinical biomarkers including PDL1 expression and tumor mutational burden (TMB) for predicting immunotherapy response and survival in patients with melanoma, lung cancer, bladder cancer, and gastric cancer.

8/ Spatial ecotypes showed powerful survival associations in 7,076 patients across 17 cancer types, and striking ability to predict immunotherapy response versus resistance pre-treatment in 1,249 patients across 4 cancer types treated with immunotherapy.

7/ We next developed, benchmarked, and applied Spatial EcoTyper to deconvolve spatial ecotype compositions from bulk RNA sequencing data.

6/ These SEs are fundamental, geographically organized units of TME biology – the parts list for the tumor microenvironment across cancer types – each with distinct cell-state assemblies.

5/ Spatial EcoTyper starts from single-cell spatial transcriptomic data. By integrating >10 million single cell and spatial transcriptomes from 132 tumors, 10 types of cancer, and 6 spatial transcriptomic modalities, we identified 9 strikingly conserved “spatial ecotypes” (SEs).

4/ In this study, we introduce Spatial EcoTyper, a machine-learning framework to systematically characterize and profile spatial cellular ecosystems in the TME, and Liquid EcoTyper, an interpretable AI approach that extends this TME profiling to the liquid biopsy setting.

3/ Multicellular ecosystems in the tumor microenvironment (TME) form dynamic signaling hubs that powerfully influence disease progression, immune evasion, and therapeutic response.