Abstract
The APOLLO 11 Project is an Italian multicenter initiative designed to transition lung cancer treatment from a standardized approach to personalized precision medicine. By integrating real-world clinical data with multi-omic biological profiles—including genomics, transcriptomics, and radiomics—through advanced Artificial Intelligence (AI), the consortium aims to develop a predictive "R-Model." This tool is intended to assist clinicians in selecting the most effective innovative therapies for individual patients based on comprehensive biological and clinical signatures.
Introduction: The Limitations of Current Biomarkers
Despite the introduction of innovative therapies such as immune checkpoint inhibitors (ICIs) and targeted treatments, lung cancer remains a leading cause of cancer-related mortality. Currently, treatment decisions often rely on a narrow set of biomarkers, most notably PD-L1 expression. However, the biological complexity of lung cancer frequently renders single-biomarker predictions inadequate, leading to suboptimal responses or "hyperprogression" in some patients.
The challenge in advancing precision oncology lies in the fragmented nature of "Big Data." Clinical, imaging, and genetic data are often siloed within individual institutions, and traditional statistical methods struggle to synthesize these heterogeneous datasets. The APOLLO 11 Project addresses these hurdles by pooling national resources and applying sophisticated machine learning techniques.
Methodology: The Three Pillars of APOLLO 11
The APOLLO 11 study utilizes both retrospective and prospective cohorts across a network of Italian oncology centers. Its methodology is structured around three core pillars:
- Data Harmonization: The project employs a distributed REDCap platform to collect and standardize clinical, biochemical, and radiological data across participating institutions. This ensures data consistency and enables large-scale analysis.
- Virtual Multilevel Biobank: Instead of a centralized physical repository, the project maintains a "virtual" biobank of multi-omic data. This includes tissue genomics, liquid biopsies (ctDNA), and immune profiling via Fluorescence Activated Cell Sorting (FACS).
- AI and Machine Learning Pipeline: The consortium utilizes a multi-stage AI workflow:
- Feature Selection: Principal Component Analysis (PCA) and Autoencoders are used to reduce dimensionality and identify the most relevant biological signals.
- Predictive Modeling: Algorithms such as Support Vector Machines (SVM), Extratrees, and Neural Networks are trained to predict treatment outcomes.
- Explainable AI (XAI): A critical component of the project is ensuring that AI-generated predictions are interpretable for clinicians, fostering trust and clinical utility.
Data Scope and the APOLLO11-BRI Cohort
The project has established a robust dataset to support its predictive modeling. The primary research focus, APOLLO11-BRI, includes over 1,000 patients with Non-Small Cell Lung Cancer (NSCLC) to identify biomarkers for immunotherapy response.
The data scope also includes:
- Validation Cohorts: 120 plasma samples for FACS-based immune profiling.
- Single-Cell Analysis: Transcriptomic data from 30 patients receiving first-line immunotherapy.
- Broad Inclusion: The study encompasses both NSCLC and Small Cell Lung Cancer (SCLC) patients treated with therapies approved since 2010.
The "R-Model" and Real-World Evidence
The ultimate goal of the consortium is the "R-Model," a decision-support tool providing individualized response predictions. By utilizing Real-World Evidence (RWE) rather than strictly controlled clinical trial populations, the model's findings are more likely to translate effectively to general clinical practice. Furthermore, the integration of radiomics allows quantitative features from medical imaging (CT/PET scans) to be correlated with molecular and clinical profiles, providing a holistic view of the patient's disease.
Current Status and Future Outlook
As a current trial report, APOLLO 11 has successfully established the necessary infrastructure for this large-scale collaboration. The consortium has integrated oncology departments, pathology units, and AI experts across Italy. The study is officially registered (NCT05550961), and the standardized data collection protocols are in place.
While technical challenges regarding data heterogeneity remain, the APOLLO 11 framework provides a scalable model for precision oncology. Future efforts will expand the scope to include specific driver mutations (e.g., EGFR, KRAS, ALK) and rarer subtypes of lung cancer, continuing the push toward truly individualized patient care.
Summary updated