In a new milestone for the DeepLife–Centre Léon Bérard collaboration, a study presented at AACR 2026 presented at AACR 2026 by our partner Pr. Pierre Saintigny aims at understanding the biological mechanisms of response to immunotherapy in patients treated with immunotherapy  for locally advanced and metastatic head and neck squamous cell carcinoma (HNSCC). Those were stratified in 3 groups with different clinical outcomes. In order to understand the single cell determinants of poor versus good prognosis, a ~660k-cells single-cell RNA sequencing oral mucosa (OMC) atlas, built in collaboration with DeepLife's, was leveraged. Building on this success, DeepLife and Centre Léon Bérard are now expanding their collaboration to extend the existing head and neck cell atlas.

‍

A signature that sees inside the tumor 

HNSCC accounts for an estimated 890,000 new cases globally each year, with locally advanced and metastatic forms carrying a poor prognosis. Immune checkpoint inhibitors have transformed treatment but only a minority of patients do benefit from them. Identifying who will benefit, and why, remains one of the field's central challenges.

A composite gene expression signature (cGES) was previously developed to improve patient stratification in a pancancer perspective : a 41-gene signature comprising 22 protective (P) and 19 adverse (A) genes allowed to stratify patients into three risk groups: low (L-risk), intermediate (I-risk), and high-risk (H-risk). Results were presented at AACR Annual Meeting in 2025 (Cancer Res 2025; 85(8_Suppl_1): 6360). The AACR 2026 poster (Cancer Res 2026; 86(7_Supplement): 1037) now delivers its validation in two prospective multicentric Unicancer trials (NCT03226756; NCT03412058) including a total of 170 patients with locally advanced and metastatic HNSCC. The results are clear and after adjustment for ECOG, age, gender, and alcohol/tobacco use, both L-risk and I-risk patients survived significantly longer without disease progression than H-risk patients (PFS hazard ratios of 0.51 and 0.64), with L-risk patients also gaining a clear overall survival advantage (OS HR 0.54).

‍

DeepLife's single-cell infrastructure 

In order to understand the biological mechanisms underlying the low- versus high-risk groups, a large-scale single-cell RNA sequencing atlas co-constructed through the DeepLife–Centre Léon Bérard partnership was leveraged, including ~660k cells from 202 samples, from 102 patients with oral cancer. In this collaboration, CLB contributed through unpublished single-cell RNA sequencing data from its own cohort of patients, while DeepLife integrated publicly available transcriptomic data from the OmicStore catalog, combining both into a unified, AI-ready resource mapping the cellular landscape of HNSCC at single-cell resolution.

Analysis of the atlas revealed a striking biological logic behind the signature. Adverse (A) genes from the cGES were predominantly expressed by epithelial and stromal cells. Protective (P) genes were expressed by immune cells, particularly T cell populations. This cell-type-level resolution, achieved through DeepLife's computational framework, translates a complex gene list into a coherent biological story. High-risk tumors emerged with an altered communication landscape, reinforced epithelial junctions and heightened EGFR signaling that is consistent with a proliferative, potentially immune-evasive phenotype. Low-risk tumors, by contrast, showed an immune-enriched network marked by adaptive immune activation and active T cell recruitment.

Spatial transcriptomics confirms the biology, spot by spot

Spatial transcriptomics, conducted independently on independent Centre Léon Bérard cases, provided a further layer of validation. Visium deconvolution on 12 patient samples confirmed what the single-cell atlas had predicted: epithelial-enriched tissue regions carried the highest adverse gene scores, while T cell-enriched regions displayed the strongest protective gene expression. H-risk and L-risk tumors showed stark, statistically significant differences across EGFR, T-cell chemotaxis, antigen presentation, and complement activation signals, thus anchoring the cellular biology of the atlas firmly in tissue-level reality.

What this means for patients, and for therapy design

By capturing the molecular logic of the tumor microenvironment, the study directly informs therapeutic strategy. High-risk patients, whose tumors are driven by EGFR-centered programs and appear to resist immune engagement, emerge as candidates for combinations targeting EGFR alongside immune checkpoints. Low-risk patients, already showing active immune mobilization, may benefit most from approaches that sustain and amplify their existing response. As we conclude, the cGES "functions not only as a prognostic biomarker but also provides mechanistic insight into disease biology" a description that equally captures what DeepLife's digital twin and cell atlas frameworks are designed to deliver at scale.

Announcing the next chapter: expanding the Head & Neck Cell Atlas

Following the presentation at AACR 2026, DeepLife and Centre Léon Bérard are proud to announce an expansion of the collaboration that made this work possible. The OMC-Atlas, originally built to map cellular trajectories in oral squamous cell carcinoma, will now be extended into a broader head and neck single-cell atlas, incorporating new patient cohorts and tumor contexts beyond the oral cavity. This next phase directly aims to create the most comprehensive single-cell reference map of head and neck cancer to date: a shared scientific asset for biomarker discovery, patient stratification, and therapeutic target identification. More details on scope, timeline, and access modalities will be shared in the coming months.

This expansion reflects a deepening commitment between the two organizations to move beyond individual studies and toward durable, large-scale scientific assets — the kind that only long-term partnership between clinical/translational expertise and AI-driven omics infrastructure can produce. More details on the scope, timeline, and access modalities will be shared in the coming months.

"Our collaboration with DeepLife has already reshaped how we understand the tumor microenvironment in head and neck cancer. Expanding the atlas is the logcal next step and we believe it will become a reference resource for the entire field."

— Dr. Pierre Saintigny, Centre Léon Bérard

‍

About the study

The results were presented at the American Association for Cancer Research (AACR) Annual Meeting 2026 by Mehdi Lamkhioued, Timothée Casini and colleagues from Centre Léon Bérard, Institut Curie, IUCT Oncopole, Gustave Roussy, Université Côte d'Azur, and DeepLife. The work was supported by two prospective multicentric clinical trials sponsored by Unicancer Group and leveraged multi-modal omics data including bulk RNA, single-cell RNA sequencing, and Visium spatial transcriptomics.

Centre Léon Bérard is a member of the Unicancer Group, which brings together a network of 20 private, not-for-profit health establishments specialised in oncology – and a key player in cancer research. It is recognized as an international center of reference in oncology. Based in Lyon - France’s second-largest city – Centre Léon Bérard performs a triple mission of care, research and education, with the constant objective to increase the quality and accessibility of innovation for patient care. For more information, visit their website.

About the team of Pierre Saintigny

Pierre Saintigny is a Medical Oncologist, Professor of Medicine at University Claude Bernard Lyon 1, Coordinator of the Department of Translational Medicine and team leader at the Cancer Research Center of Lyon. His team - Cancer dynamics, adaptation, and modeling (CASTING) - focuses on the global biological changes occurring during oral tumorigenesis. The team integrates clinical, pathological, cellular and molecular profiles both in human samples and preclinical models to unravel the dynamic changes during early stages of tumorigenesis and under the selective pressure of therapy, to reveal vulnerabilities that may in turn translate into new treatment and prevention strategies.

contacts

kevin.carvalho@deeplife.co

‍