Identification of Novel Diagnostic Biomarkers and Host-Directed Drug Screening for Mycobacterium avium Infection: A Multi-Omics and Artificial Intelligence Study.

OBJECTIVE: The rising incidence of Mycobacterium avium (MAV) infection poses significant challenges due to diagnostic delays and refractory treatment. Understanding host immune remodeling-particularly myeloid heterogeneity-is critical for identifying precise therapeutic targets. METHODS: We integrated single-cell RNA sequencing (scRNA-seq) with machine learning to profile peripheral blood from MAV patients. CellChat analysis mapped intercellular communication, while an ensemble of 113 machine learning algorithms screened for diagnostic biomarkers. Network pharmacology and molecular docking were employed to identify host-directed therapies, validated via RT-qPCR. RESULTS: The single-cell atlas revealed a pathological expansion of hyper-inflammatory activated macrophages and neutrophils in progressive disease. Analysis identified a monocyte-derived MIF signaling axis that recruits and "locks" APP-expressing macrophages, driving immune dysregulation. A diagnostic model (Stepglm+GBM) based on five core genes (IL1B, STAT3, TNF, STAT1, TLR4) achieved high accuracy (AUC > 0.88). Furthermore, molecular docking confirmed that Wogonin forms high-affinity bonds with STAT3 and TNF, suggesting its potential as an immunomodulator. CONCLUSION: This study delineates the high-resolution immune landscape of MAV infection, elucidating the MIF-APP axis-driven myeloid reprogramming. We established a robust diagnostic model and identified Wogonin as a potential therapeutic agent, providing novel strategies to break the "immune stalemate" in MAV infection.