# Integrative analyses of metastatic cancer transcriptome reveal clinically distinct cellular States and ecosystems

## TL;DR

This study identifies distinct cellular states and ecosystems in metastatic cancer, offering insights into cancer diversity and potential treatment targets.

## Contribution

The study introduces a machine learning framework to characterize metastatic cancer ecosystems and their clinical relevance.

## Key findings

- Identified 45 distinct cellular states across 12 cell types in metastatic cancer samples.
- Discovered five ecotypes linked to different clinical outcomes and key transcription factors associated with patient survival.
- Marker genes of cellular states were enriched in cancer hallmark and immune-related pathways.

## Abstract

Determining the diverse cellular states and their organization into cellular ecosystems that make up metastatic tumor is vital for elucidating the biological and prognostic diversity of cancer. However, large-scale studies profiling the clinical relevance of these cellular states and ecotypes are still lacking in metastatic cancers. In this study, we used EcoTyper, a machine learning framework, to comprehensively analyze transcriptomes from 2822 metastatic cancer patient samples covering 25 cancer types, enabling characterization of the fundamental cellular states and tumor ecosystems integral to metastatic cancer. We identified 45 distinct cellular states across 12 cell types and validated their robustness in validation cohorts. We observed that they differed in functional and prognostic associations. Survival analysis revealed that the clinically relevant cellular states, highlighting their promise as predictors of clinical outcomes. Functional enrichment analysis exhibited that the marker genes of cellular states were significantly enriched in cancer hallmark and immune-related pathways. In addition, our analysis identified five ecotypes associated with different clinical outcomes. Transcription factor enrichment analysis revealed key transcription factors (i.e. SPIB, SRF, and NR1D1) that were significantly associated with patient clinical outcomes. In conclusion, this study provided a high-resolution landscape of cellular states and ecosystems in metastatic tumors, offering new potential targets for the development of cancer treatment strategies and prognostic assessment.

The online version contains supplementary material available at 10.1038/s41598-026-36512-3.