Related Experiment Video
Updated: Jun 21, 2026
Shear Assay Protocol for the Determination of Single-Cell Material Properties
Published on: May 19, 2023
Machine learning-driven QSAR modeling combined with single cell transcriptomics identifies novel drug targets for
Nagasundaram Nagarajan1, Sushil Kumar Shakyawar1, Kayode Raheem1
1Department of Genetics, Cell Biology and Anatomy, University of Nebraska Medical Center, Omaha, NE, USA.
Background:
Non-small cell lung cancer (NSCLC) is a leading cause of cancer-related mortality, largely due to frequent metastasis to the brain and bones. Therapeutic outcomes are often limited by drug resistance, tumor heterogeneity, and the lack of effective treatment options across different stages and metastatic sites. Identifying druggable targets that are conserved between primary tumors and metastases is critical for advancing precision oncology.
Methods:
scRNA-seq expression profiles from primary NSCLC tumors and matched brain and bone metastases were analyzed to identify conserved and site-specific gene expression signatures. Ingenuity Pathway Analysis was used for target prioritization. Potential targets identified from scRNA-seq analysis were used for ligand screening using machine learning (ML)-based quantitative structure-activity relationship (QSAR) modeling. QSAR models were developed using ChEMBL bioactivity data and evaluated across multiple ML algorithms. Large-scale virtual screening was followed by molecular docking and molecular dynamics (MD) simulations for lead optimization.
Results:
Eight candidate therapeutic targets were prioritized, among which ARPC2, PSMB4, and RAC2 were consistently overexpressed across primary, brain, and bone metastatic sites and were functionally implicated in key cancer-associated pathways. QSAR modeling demonstrated strong predictive performance, with XGBoost and Random Forest models achieving AUROC values greater than 0.97. Virtual screening of approximately 9-15 million compounds per target identified high-affinity candidates. Subsequent docking and MD simulations revealed that the ARPC2-14465616, PSMB4-74833722, and RAC2-57175325 complexes exhibited the highest structural stability and sustained intermolecular interactions.
Conclusion:
This integrative single-cell transcriptomics and ML-driven drug discovery framework identified conserved druggable targets and promising lead compounds for metastatic NSCLC. The results provide a strong foundation for experimental validation and the development of novel therapeutic strategies targeting both primary tumors and metastatic lesions.
Insights
This study identifies conserved druggable targets like ARPC2, PSMB4, and RAC2 in non-small cell lung cancer (NSCLC) metastases. Machine learning identified promising drug candidates for treating primary tumors and metastatic lesions.
Area of Science:
- Oncology
- Computational Biology
- Drug Discovery
Background:
- Non-small cell lung cancer (NSCLC) is a major cause of cancer mortality, with frequent metastasis to the brain and bones.
- Limited treatment options and drug resistance in metastatic NSCLC necessitate novel therapeutic targets.
- Identifying conserved druggable targets across primary and metastatic sites is crucial for precision oncology.
Purpose of the Study:
- To identify conserved druggable targets in metastatic non-small cell lung cancer (NSCLC).
- To discover novel lead compounds for NSCLC using a machine learning-driven drug discovery framework.
- To provide a foundation for developing new therapies for both primary and metastatic NSCLC.
Main Methods:
- Single-cell RNA sequencing (scRNA-seq) analyzed gene expression in primary NSCLC and matched metastases.
- Ingenuity Pathway Analysis prioritized therapeutic targets.
- Machine learning (ML) models, including quantitative structure-activity relationship (QSAR), screened for potential drug ligands.
- Virtual screening, molecular docking, and molecular dynamics simulations optimized lead compounds.
Main Results:
- ARPC2, PSMB4, and RAC2 were identified as consistently overexpressed targets in primary and metastatic NSCLC.
- ML-based QSAR models achieved high predictive performance (AUROC > 0.97).
- Virtual screening identified high-affinity drug candidates, with subsequent simulations confirming structural stability and interactions for ARPC2, PSMB4, and RAC2 complexes.
Conclusions:
- An integrated approach using single-cell transcriptomics and ML identified conserved druggable targets for metastatic NSCLC.
- Promising lead compounds were discovered, offering potential for novel therapeutic strategies.
- The findings support further experimental validation for treating primary tumors and metastatic lesions.
