Head and neck squamous cell carcinoma (HNSCC) exhibits considerable variability in patient outcome. It has been reported that SOX2 plays a role in proliferation, tumor growth, drug resistance, and metastasis in a variety of cancer types. Additionally, SOX9 has been implicated in immune tolerance and treatment failures. SOX2 and SOX9 induce treatment failure by a molecular mechanism that has not yet been elucidated. This study explores the inverse association of SOX2/SOX9 and their distinct expression in tumors, influencing the tumor microenvironment and radiotherapy responses. Through public RNA sequencing data, human biopsy samples, and knockdown cellular models, we explored the effects of inverted SOX2 and SOX9 expression. We found that patients expressing SOX2^LowSOX9^High showed decreased survival compared to SOX2^HighSOX9^Low. A survival analysis of patients stratified by radiotherapy and human papillomavirus brings additional clinical relevance. We identified a gene set signature comprising newly discovered candidate genes resulting from inverted SOX2/SOX9 expression. Moreover, the TGF-β pathway emerges as a significant predicted contributor to the overexpression of these candidate genes. In vitro findings reveal that silencing SOX2 enhances tumor radioresistance, while SOX9 silencing enhances radiosensitivity. These discoveries lay the groundwork for further studies on the therapeutic potential of transcription factors in optimizing HNSCC treatment.

Multidisciplinary tumor boards integrate longitudinal treatment histories, molecular profiling and rapidly evolving evidence to guide decisions in hematological malignancies, yet access to this level of subspecialty deliberation is increasingly uneven. Here we develop HemaGuide, a locally deployable, modular large language model agent that converts unstructured clinical documents into structured case representations, autonomously routes cases to specialized decision modes (‘guideline’, ‘advanced’ and ‘molecular’) and grounds recommendations in disease-specific guideline flowcharts and a clinical decision memory of >2,000 real-world tumor board cases. In expert-blinded benchmarking on 45 high-complexity cases across six foundation models, HemaGuide substantially improved concordance with tumor board decisions. A systematic ablation study across 11 layers confirmed that performance gains were routing-type-dependent, with no single component sufficient across case types. Automated classification of 70 clinically relevant missense variants showed high concordance with expert standards; no oncogenic variant was downgraded to benign and the whole workflow was completed under real-time conditions on commodity hardware with a median latency of 39 s rather than the hours typically required for manual molecular board workflows. In a simulated practice study, agent-assisted resident physicians achieved near-senior concordance and partially outperformed senior physicians in their subspecialty. External validation on 555 independent cases from a second academic center yielded 81.8% concordance across 47 entities, and a prospective 1-month silent trial on 64 consecutive, unselected cases achieved 82.8% concordance. Hallucinations occurred in 2 of 664 evaluated cases (0.3%). Together these data provide evidence that locally deployable, case-grounded large language model agents can deliver auditable clinical decision support across hematological malignancies, with concordance maintained across institutions and under real-time conditions on commodity hardware.