Development of a multi-indicator risk prediction model for cervical cancer associated with benzo[a]pyrene and nicotine exposure: A multi-omics study integrating toxicological analyses and molecular docking.

Background: Exposure to the tobacco-related compounds Benzo[a]pyrene and Nicotine has been associated with the development of several diseases. The aim of this study was to investigate the common genes associated with cervical cancer, construct a risk prediction model to reveal their biological functions, and evaluate the prognostic significance of the model to identify its potential value in the treatment of cervical cancer.

Methods: In this study, genes associated with Benzo[a]pyrene and Nicotine and cervical cancer-related genes were screened by multiple databases. Target genes were analysed using a multi-omics machine learning algorithm to construct a risk-prognostic model, and nine key target genes were identified. The risk prediction models were evaluated by univariate and multivariate Cox regression analyses, and model validation was performed using the TCGA and GSE44001 datasets. In addition, clinical relevance, biofunctional enrichment, immune infiltration, and drug sensitivity analyses were performed, and the binding affinities of the two compounds to the target genes were investigated by combining molecular docking and kinetic analyses, and the Mendelian randomisation method was applied to analyse the causal association between the target genes and cervical cancer.

Results: A total of 682 genes associated with the two compounds were screened by ChEMBL, STITCH and SwissTargetPrediction databases, while 1451 genes associated with cervical cancer were identified by using GeneCards and OMIM databases, among which 109 genes were associated with both the two compounds and cervical cancer. The degree of interaction between different genes was determined by protein interaction network analysis. Based on various machine learning algorithms, a risk prediction model associated with Benzo[a]pyrene and Nicotine exposure and cervical cancer was constructed, and the good prediction performance of the model was verified in TCGA and GSE44001 datasets. In addition, a column-line diagram associated with the risk prediction model was constructed to provide a clinical tool for predicting prognosis. Further analyses revealed significant differences in the enrichment of biological processes, immune-infiltrating cells and immunomodulatory factors between the high-risk and low-risk groups, and the risk prediction model was strongly correlated with drug susceptibility, showing significant associations especially in tipifarnib-P1, AZD3463, docetaxel and AT-7519. Molecular docking and molecular dynamics simulations revealed a strong binding affinity between Benzo[a]pyrene and SLAMF6. Furthermore, Mendelian randomisation analysis revealed a significant causal relationship between SLAMF6 and AIG1.

Conclusions: Risk prediction models based on multi-omics data and machine learning algorithms provide potential reference targets for prognosis prediction and personalised treatment of cervical cancer patients. The results of this study provide important insights into the understanding of the health risks of cervical cancer associated with Benzo[a]pyrene and Nicotine exposures and the development of preventive and therapeutic strategies for cervical cancer, which may contribute to the development of precision medicine for cervical cancer.