Multi-Omics Inflammatory Phenotype for ABPA Recurrence Risk Prediction

Multi-Omics Data-Derived Inflammatory Phenotype for ABPA Recurrence Risk Prediction: A Multicenter Study

To develop and externally validate a machine learning model for predicting the 1-year risk of relapse in patients with stable ABPA, and to further evaluate its value in risk stratification and clinical decision-making.

This project aims to develop an inflammatory phenotype-based risk prediction model for recurrence of allergic bronchopulmonary aspergillosis (ABPA) to enable stratified patient management. The study integrates multidimensional data sources, including radiomics, mycobiomics, inflammatory biomarkers, pulmonary function parameters, and routine clinical records. Deep machine learning algorithms are employed to extract and select key features from these multi-omics and clinical datasets, define inflammatory phenotypes, and subsequently construct a recurrence risk prediction model. Based on the risk stratification derived from the model, low-risk individuals will receive regular follow-up, whereas high-risk individuals will undergo intensified intervention and management. This approach is expected to optimize individualized treatment strategies for ABPA patients, reduce recurrence rates, and improve clinical outcomes.

Inclusion Criteria:

• Female and Male patients aged 18-80 years
• diagnosis of Allergic Bronchopulmonary Aspergillosis ABPA accroding to the 2024 ISHAM Working Group Diagnostic Criteria

Exclusion Criteria:

• Patients with malignant tumors or severe organ dysfunction (e.g., cardiac, cerebral, renal, etc.)
• Patients with severe comorbidities, including active pulmonary tuberculosis, lung cancer, chronic heart failure (NYHA class Ⅳ), chronic kidney disease (CKD stage 5), decompensated cirrhosis, etc.
• Patients with immunosuppressive status, such as HIV infection, long-term use of oral corticosteroids or immunosuppressive agents.
• Pregnant or lactating women.
• Patients with missing key data or incomplete medical records.