# Machine learning-driven identification of shared and disease-specific mitochondria-related genes in COPD, NSCLC, and NSCLC with COPD

## TL;DR

This study identifies blood-based mitochondrial genes that can distinguish COPD, NSCLC, and their coexistence, offering potential diagnostic and treatment options.

## Contribution

The study introduces novel mitochondrial gene signatures and drug repurposing candidates for COPD, NSCLC, and their comorbidity.

## Key findings

- Mitochondrial gene signatures (NDUFB6/BID/COX7A2) achieved high diagnostic accuracy (AUC >0.7) across COPD, NSCLC, and their overlap.
- Machine learning models revealed mitochondria-immune crosstalk in disease progression and immune cell composition changes.
- Metformin and ME-344 were identified as potential repurposed drugs targeting mitochondrial genes in these conditions.

## Abstract

Chronic obstructive pulmonary disease (COPD) and non-small-cell lung cancer (NSCLC) often coexist; here, the shared mitochondrial drivers were investigated. Serum from 30 subjects (seven controls, nine COPD, eight NSCLC, and six NSCLC with COPD) underwent RNA-seq, integrated with 1,136 MitoCarta 3.0-derived mitochondrial-related genes (MRGs). DESeq2 identified 25, 124, and 58 mitochondria-related differentially expressed genes (MR-DEGs) in COPD, NSCLC, and their comorbidity, respectively, with 15 and 58 overlapping genes in relevant pairs. SVM-RFE selected two biomarker sets (3-gene and 5-gene), showing excellent diagnostic performance via ROC (AUC 0.89–0.92) and accurate multivariate logistic regression models. GSEA highlighted immune-inflammatory and oxidative phosphorylation pathways; CIBERSORT revealed altered immune cell proportions (e.g., elevated monocytes in COPD) with biomarker-immune cell correlations. CTD linked BID/COX7A2 to NSCLC, and DGIdb identified metformin/ME-344 as potential drugs. These mitochondrial gene signatures, validated in blood as robust classifiers of COPD, NSCLC, and their overlap, simultaneously furnish diagnostic biomarkers and actionable therapeutic targets, underscoring the translational value of mitochondria-immune crosstalk.

•Blood-based mitochondrial gene signatures distinguish COPD, NSCLC and NSCLC with COPD•NDUFB6/BID/COX7A2 yield ROC-AUC >0.7 across all groups; validated by qPCR•Machine-learning models reveal mitochondria-immune crosstalk in COPD, NSCLC & NSCLC with COPD•Metformin & ME-344 nominated to target NDUFB6/NDUFA1 via drug-repurposing

Blood-based mitochondrial gene signatures distinguish COPD, NSCLC and NSCLC with COPD

NDUFB6/BID/COX7A2 yield ROC-AUC >0.7 across all groups; validated by qPCR

Machine-learning models reveal mitochondria-immune crosstalk in COPD, NSCLC & NSCLC with COPD

Metformin & ME-344 nominated to target NDUFB6/NDUFA1 via drug-repurposing

Health sciences; Medicine; Medical specialty; Internal medicine; Oncology; Respiratory medicine