{
  "id": 14232697,
  "title": "Deep-learning-driven tumor microenvironment profiling improves immunotherapy response prediction",
  "url": "https://urgent.news/2026/10/13/deep-learning-driven-tumor-microenvironment-profiling-improves",
  "topic": "health",
  "section": "Health & Medicine",
  "published": "2026-10-13T00:00:00.000Z",
  "source": {
    "name": "bioRxiv",
    "slug": "biorxiv",
    "url": "https://www.biorxiv.org/content/10.64898/2026.10.06.757143v1?rss=1"
  },
  "original_language": "en",
  "account": "Immunotherapy has proven effective in treating many cancers, but only works for a fraction of patients. The reason for this variability lies in the tumor microenvironment (TME), which is influenced by complex and interconnected signaling networks. Existing methods for predicting immunotherapy response often rely on isolated cell types or signatures, missing the bigger picture of the signaling network within the TME.\n\nA new computational framework called EcoNet has been developed to tackle this issue. EcoNet begins by identifying co-occurring cell states, then constructs the corresponding intercellular and intracellular signaling networks. By applying a graph attention network to these networks, EcoNet generates embeddings that can predict immunotherapy responses.\n\nEcoNet was trained on 2,532 pan-cancer immunotherapy samples from various sources. Compared to existing methods, EcoNet showed a significant improvement in accuracy, boosting the area under the curve (AUC) by 7.5-19.9% and marker-based signatures by 4.4-21.2%. Additionally, the genes EcoNet focused on, termed high-attention genes, were enriched in those crucial for T-cell-mediated tumor killing, as verified by an independent CRISPR screen. When applied to a ccRCC cohort, EcoNet successfully stratified the patients based on their clinical outcomes (P=0.024).\n\nThese findings were further validated using CosMx single-cell spatial transcriptomics data from a ccRCC clinical cohort of 46 samples, containing approximately 450,000 cells. The TME configuration predicted by EcoNet was found to be localized in immune-enriched niches, confirming its predictive power.\n\nIn conclusion, EcoNet represents a generalizable approach for predicting therapeutic response based on TME characteristics. Its potential application extends beyond ccRCC to other cancers and therapeutic contexts.",
  "summary": "Immunotherapy has transformed the management of many cancers, such as clear cell renal cell carcinoma (ccRCC), yet durable clinical benefit is achieved in only a subset of patients. Immunotherapy outcomes are governed by the multicellular tumor microenvironment (TME), through the structured and directional interplay of regulatory programs and ligand-receptor communications. Yet most existing…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}