Integrated Multi-Omics Analysis Identifies Key Genes and Drug Candidates in Non-Small Cell Lung Cancer

Technical Summary: Integrated Multi-Omics Analysis Identifies Key Genes and Drug Candidates in Non-Small Cell Lung Cancer

Problem Statement
Non-small cell lung cancer (NSCLC) accounts for approximately 85% of lung cancer cases but remains associated with a poor prognosis, with a 5-year survival rate below 20%. This poor outcome is attributed to late diagnosis, tumor heterogeneity, and the complexity of the tumor microenvironment (TME). Current therapeutic strategies, including surgery, radiotherapy, and chemotherapy, often face limitations due to drug resistance and lack of specificity. Consequently, there is an urgent clinical need to identify highly sensitive and specific early diagnostic markers and novel targeted therapeutic agents. While traditional differential expression analysis offers initial insights, it often lacks the ability to capture co-expression relationships and network regulations inherent in tumorigenesis. Furthermore, the specific cellular distribution of key oncogenes within the NSCLC microenvironment and their potential as precise drug targets require deeper investigation.

Methodology
This study employed a comprehensive multi-omics integration strategy combining bulk transcriptomics, single-cell RNA sequencing (scRNA-seq), and computational drug discovery approaches:

1. Data Acquisition and Preprocessing: Transcriptomic data from six GEO datasets (GSE19804, GSE75037, GSE102287, GSE103512, GSE116959, GSE118370) containing NSCLC tumor and normal tissue samples were merged. Batch effects were corrected using the ComBat algorithm, and low-expression genes were filtered.
2. Gene Screening Pipeline:
   • Differential Expression Analysis (DEA): Identified differentially expressed genes (DEGs) using the limma package (criteria: |log2FC| > 1, adj. P < 0.05).
   • Weighted Gene Co-expression Network Analysis (WGCNA): Constructed a co-expression network to identify modules correlated with clinical traits. The module most positively correlated with NSCLC was selected.
   • Intersection and Enrichment: The intersection of DEGs and the key WGCNA module genes formed the core gene set, followed by GO and KEGG functional enrichment analysis.
   • PPI Network: Protein-protein interaction (PPI) networks were constructed via the STRING database, and hub genes were identified using the MCODE and cytoHubba plugins.
   • Machine Learning Integration: Three algorithms—LASSO regression, Support Vector Machine with Recursive Feature Elimination (SVM-RFE), and Random Forest (RF)—were applied to the hub genes. The intersection of genes selected by all three methods defined the final key gene set.
3. Diagnostic Modeling: Receiver Operating Characteristic (ROC) curves and a nomogram prediction model were constructed to evaluate diagnostic performance and clinical utility.
4. Single-Cell Analysis: Public scRNA-seq data (GSE229253) were analyzed using the Seurat package for cell clustering and annotation. The CellChat package was utilized to map cell-cell communication networks and ligand-receptor interactions within the TME.
5. Virtual Knockout and Drug Screening:
   • Virtual Knockout: The scTenifoldKnk tool was used to predict the downstream effects of TOP2A inactivation on the gene expression network.
   • Virtual Screening: The DrugReflector deep learning framework screened for small-molecule drugs targeting TOP2A.
   • Molecular Docking: Candidate drugs were docked against the TOP2A protein structure using CB-Dock2 to calculate binding energies.

Key Contributions and Results

• Identification of Key Diagnostic Genes: The study identified 1,159 DEGs and a key WGCNA module (MEblue) containing 758 genes. The intersection yielded 111 genes, from which 10 hub genes were selected via PPI analysis. Machine learning integration (LASSO, SVM, RF) refined this to a final set of seven key genes: CCNA2, CDC20, BUB1, CCNB1, TOP2A, BIRC5, and CENPF.
• Diagnostic Performance: All seven genes demonstrated excellent diagnostic value with Area Under the Curve (AUC) values > 0.85. TOP2A showed the highest efficacy (AUC = 0.957). A nomogram based on these genes showed good calibration and provided a positive net benefit in decision curve analysis.
• Prognostic Significance: High expression of TOP2A was significantly associated with poor overall survival (HR = 1.48, P = 1.5×10⁻¹⁰) and was upregulated across multiple cancer types.
• Single-Cell Insights: Analysis revealed that the NSCLC microenvironment is dominated by epithelial cells, which exhibit high expression of the seven key genes, particularly TOP2A. Cell-cell communication analysis identified critical ligand-receptor pathways involving TOP2A-high epithelial cells, including VEGFA-VEGFR1 (angiogenesis), HGF-MET (invasion/metastasis), and GAS6-AXL (immune evasion).
• Mechanistic Discovery (Virtual Knockout): Virtual knockout of TOP2A significantly altered the expression of EPS8, a key adaptor protein in the EGFR pathway involved in cytoskeleton reorganization and cell migration. This suggests a potential "TOP2A-EPS8-EGFR/actin" axis in promoting metastasis.
• Drug Candidate Identification: Virtual screening identified BRD-K59184148 as the top candidate drug targeting TOP2A. Molecular docking confirmed a stable complex with a binding energy of -9.8 kcal/mol, involving hydrogen bonds and hydrophobic interactions with key residues (e.g., ARG-804, SER-802).

Significance and Claims
The authors claim that this study provides a new strategy for NSCLC diagnosis and treatment through the following contributions:

1. Novel Biomarkers: The seven identified genes serve as excellent diagnostic markers for NSCLC, with TOP2A standing out as a specific, high-efficacy marker highly expressed in epithelial cells.
2. Microenvironment Elucidation: The study elucidates the role of TOP2A-high epithelial cells as central signal senders in the TME, influencing angiogenesis, invasion, and immune evasion through specific paracrine signaling pathways.
3. Therapeutic Targeting: By linking TOP2A to EPS8, the study offers a new mechanistic clue regarding the pro-metastatic nature of TOP2A.
4. Drug Discovery: The identification of BRD-K59184148 as a potential lead compound with strong binding affinity to TOP2A offers a new avenue for developing selective inhibitors, potentially overcoming the toxicity and resistance associated with existing TOP2A inhibitors like doxorubicin and etoposide.

The authors maintain a modest stance, acknowledging that these findings are based on public database analysis and require prospective clinical validation, protein-level verification, and in vivo/in vitro efficacy studies before clinical translation.