Identification of neuronal synapse-related signatures and potential therapeutic drugs in colorectal cancer based on machine learning algorithms and molecular docking.

Background: Nervous system-cancer interactions can regulate tumorigenesis, invasion, and metastasis. However, specific biomarkers for targeting neuron synapse in colorectal cancer (CRC) remain unexplored. This study aims to develop a neuronal synapse-related signature (NSRS) to predict survival in CRC patients, identify potential therapeutic drugs, and explore its clinical applications.

Methods: We collected neuronal synapse genes (NSGs) from the Molecular Signatures Database (MSigDB) and published mass spectrometry data. Using weighted gene co-expression network analysis (WGCNA) and least absolute shrinkage and selection operator Cox regression (LASSO-Cox), we identified prognostic NSGs and constructed a NSRS through multivariate Cox regression. Functional enrichment analysis revealed the molecular characteristics of NSRS subgroups. Additionally, xCell and ESTIMATE algorithms quantified the abundance of 54 cell subtypes and assessed the tumor immune microenvironment (TIME) of the two NSRS subgroups. Finally, drug prediction and molecular docking identified candidate drugs with therapeutic potential.

Results: Seven key prognostic NSGs were identified, and an independent, stable NSRS model was constructed. Kaplan-Meier survival curves indicated that the high NSRS group had poorer outcomes (log-rank test, P<0.05). Functional enrichment analysis revealed significant enrichment of epithelial-mesenchymal transition, hypoxia, and inflammation features in the high NSRS group. xCell and ESTIMATE analyses showed a more complex TIME and lower tumor purity in the high NSRS group, highlighting the role of neuro-tumor interactions in CRC. Drug prediction and molecular docking suggested alprostadil, dihydroergocristine, and nocodazole as candidate drugs for CRC treatment.

Conclusions: This is the first study to develop neuron synapse-related biomarkers from the perspective of neuron-cancer interactions using machine learning. We constructed a robust NSRS model and identified candidate drugs targeting prognostic NSGs, providing new insights into CRC prognosis and treatment.