Machine-Learning Analysis Identifies Nine Conserved Spatial Ecotypes Across Human Tumors

A machine-learning framework for multi-analyte profiling of spatially dependent cell states and multicellular ecosystems, termed spatial ecotypes (SEs), was presented in research published in the journal Nature. The framework integrates over 10 million single-cell and spot-level spatial transcriptomes from diverse human carcinomas and melanomas.

From that dataset, researchers identified nine spatial ecotypes with broad conservation across tumor types.

Each of the nine SEs has unique biology, geospatial features and clinical outcome associations. Several SEs are linked to immunotherapy response. " Its subjects include Biomarkers, Cancer microenvironment, Genomics, Immunotherapy and Machine learning.

Multicellular programs in the tumour microenvironment drive cancer pathogenesis and response to therapy but remain challenging to identify and profile clinically. The new framework addresses that gap by defining reproducible multicellular units. SEs were distinguishable by DNA methylation profiling.

They were also recoverable from plasma cell-free DNA using deep learning. In an analysis of cfDNA from nearly 100 patients with melanoma, SE levels exhibited striking associations with immunotherapy response. The findings point to a potential route for non-invasive monitoring of the tumour microenvironment.

A related paper titled "Multiplexed imaging mass cytometry reveals distinct tumor-immune microenvironments linked to immunotherapy responses in melanoma" was published open access on 21 October 2022. The data reveal fundamental units of tumour microenvironment organization. They also demonstrate a multimodal platform for profiling both solid and liquid tumour microenvironments.

Such a platform carries implications for improved risk stratification and therapy personalization in cancer care.