Artificial intelligence applications in biomedicine face major challenges from data privacy requirements. To address this issue for clinically annotated tissue proteomic data, we developed a federated deep learning approach (ProCanFDL), training local models on simulated sites containing data from a pan-cancer cohort (n = 1,260) and 29 cohorts held behind private firewalls (n = 6,265), representing 19,930 replicate data-independent acquisition mass spectrometry runs. Local parameter updates were aggregated to build the global model, achieving a 43% performance gain on the hold-out test set (n = 625) in 14 cancer subtyping tasks compared with local models and matching centralized model performance. The approach’s generalizability was demonstrated by retraining the global model with data from two external, data-independent acquisition mass spectrometry cohorts (n = 55) and eight acquired by tandem mass tag proteomics (n = 832). ProCanFDL presents a solution for internationally collaborative machine learning initiatives using proteomic data, for example, for discovering predictive biomarkers or treatment targets while maintaining data privacy. Significance: A federated deep learning approach applied to human proteomic data, acquired using two distinct proteomic technologies from 40 tumor cohorts across eight countries, enabled accurate cancer histopathologic subtyping while preserving data privacy. This approach will enable the privacy-compliant development of large-scale proteomic artificial intelligence models, including foundation models, across institutions globally.

Background: Carcinoma of unknown primary (CUP) remains an important tumor entity and a disproportionate cause of cancer mortality. Little is known about the contemporary clinical characteristics, treatment patterns, and outcomes of CUP patients based on updated international classification guidelines. We evaluated a contemporary CUP cohort to provide insight into current clinical practice and the impact of tissue of origin assignment, site‐specific and empirical therapy in a real‐world setting.Methods: We conducted a retrospective cohort study of CUP patients, as defined by the updated European Society of Medical Oncology (ESMO) 2023 guidelines, across three tertiary referral centers in Australia between 2015 and 2022. We analyzed clinical characteristics, treatment patterns, and survival outcomes using the Kaplan–Meier method and Cox regression proportional hazard model between favorable and unfavorable risk groups.Results: We identified a total of 123 CUP patients (n = 86 unfavorable, n = 37 favorable risk as per the 2023 ESMO guidelines). Sixty‐four patients (52%) were assigned a tissue of origin by the treating clinician. Median progression free survival (PFS) was 6.8 (95% confidence interval (CI) 5.1–12.1) months and overall survival (OS) 10.2 (95% CI 6.0–18.5) months. Unfavorable risk (hazard ratio [HR] 2.9, p = 0.006), poor performance status (HR 2.8, p < 0.001), and non‐squamous histology (HR 2.5, p < 0.05) were associated with poor survival outcome. A total of 70 patients (57%) proceeded to systemic therapy. In patients with non‐squamous histology and unfavorable risk, site‐specific therapy compared to empirical chemotherapy did not improve outcome (median OS 8.2 vs. 11.8 months, p = 0.7).Conclusions: In this real‐world cohort, CUP presentations were heterogenous. Overall survival and rates of systemic treatment were poor. Poor performance status and unfavorable risk were associated with worse survival. For most patients, site‐specific therapy did not improve survival outcome. Improved and timely access to diagnostic tests and therapeutics for this group of patients is urgently required.

Background: Durvalumab following concurrent chemoradiotherapy is standard treatment for unresectable stage III non‐small‐cell lung cancer based on the results of the PACIFIC trial. Based on trial criteria, not all patients are eligible for durvalumab in routine clinical practice.Methods: We evaluated eligibility for durvalumab in a real‐world clinical setting and the impact of eligibility on outcomes. Consecutive patients treated with concurrent chemoradiotherapy at two tertiary centers between January 2015 and June 2022 were assessed. Clinical characteristics and outcomes were evaluated based on eligibility criteria for the PACIFIC trial.Results: A total of 126 patients were included. Seventy patients (56%) were eligible for durvalumab. Ineligibility was associated with shorter progression‐free survival of 9.7 months versus 18.4 months (hazard ratio [HR] 0.61, 95% confidence interval [CI] 0.39–0.95, p = 0.029) and overall survival of 26.4 months versus 58.7 months (HR 0.47, 95% CI 0.28–0.80, p = 0.005). Common reasons for ineligibility were history of previous malignancy (32%) and progressive disease or death during chemoradiotherapy (25%). Ineligible patients who received durvalumab had similar outcomes to eligible patients who received durvalumab.Conclusions: In a real‐world cohort, adjuvant durvalumab is safe and beneficial in a substantial proportion of patients who would not have been eligible for the PACIFIC trial.

Per class AUROC using ProCan Compendium and external datasets