The core of the project has been the execution of the High-Definition Oncology (HDO) study, a prospective, multicenter observational study designed to deeply profile patients with metastatic cancer. The protocol and its feasibility have been established and are detailed in the manuscript "Longitudinal clinical, physiological and molecular profiling of female metastatic cancer patients: protocol and feasibility of a multicenter high-definition oncology study," currently in its second revision at JCO Precision Oncology. This foundational work outlines the methodology for longitudinally collecting data across eleven distinct modalities, including genomics, epigenomics, microbiome, metabolome, proteome, continuous physiological monitoring via wearables, and patient-reported outcomes (PROs). Feasibility data from the initial cohort of patients demonstrated high compliance and data quality, with patient engagement in clinical visits, sample submission, and digital monitoring exceeding expectations. This success was communicated at the ESMO congress in the work "Constructing a high-definition patient-digital twin (PDT) in treatment-naïve women with advanced cancer" (2024), highlighting the viability of constructing high-definition 'patient-digital twins'.
A significant achievement has been the development of an explainable AI model for cancer prognosis that leverages PROs, presented at the ESMO congress in the communication "Explainable AI model for cancer prognosis using weighted analysis of patient-reported outcomes (PROs)" (2025). By applying a sparse, interpretable neural network to item-level data from the EORTC QLQ-C30 and GHQ-28 questionnaires, we can predict progressive disease within 12 months with high accuracy (AUC ≈ 0.85). This model surpasses traditional scoring methods by identifying and weighting the most predictive symptoms and functional items, demonstrating the prognostic power of longitudinally collected PROs.
Another key technical advance, communicated at the ASCO meeting in "Remote physiologic and behavioral monitoring to predict early treatment response in metastatic cancer: High-Definition Oncology study (HDOs) preliminary results" (2026), involves the use of remote monitoring data to predict treatment response. Analysis of passively collected data from smartwatches (e.g. heart rate, sleep, activity) and actively reported behavioral data (e.g. emotions) revealed strong associations with clinical benefit at the first tumor assessment. We identified specific 'Response-Associated Behavioral Patterns' (RABPs) that can flag patients at high risk of early disease progression. This work has culminated in a submitted manuscript describing a novel alert system based on this behavioral data, designed to detect disease progression earlier.
Finally, the project has driven significant technical innovations in the analysis of epigenetic data. The publication "Applicability of epigenetic age models to next-generation methylation arrays" (Garma & Quintela-Fandino, 2024) in Genome Medicine addresses the critical issue of cross-platform compatibility for epigenetic clocks. We developed and validated a new epigenetic age model that is compatible with both legacy and next-generation DNA methylation arrays (EPICv2), ensuring the continued utility of established biomarkers. Furthermore, the pre-print manuscript "A breast tissue-specific epigenetic clock provides accurate chronological age predictions and reveals de-correlation of age and DNA methylation in tumor-adjacent and tumor samples" describes the creation of a novel Breast Tissue-specific Epigenetic Clock (BTEC). This clock provides far more accurate age predictions in breast tissue than general pan-tissue models and has revealed a de-correlation between chronological age and DNA methylation in tumor and tumor-adjacent tissues, with tumors appearing epigenetically 'younger', particularly in more aggressive subtypes.
