Molecular docking insights into miR-155 and VEGF synergy: colorectal cancer detection through AI-enhanced integration of molecular biomarkers and clinical risk assessment

Abstract

Colorectal cancer (CRC) remains a leading cause of cancer-related mortality worldwide, underscoring the need for non-invasive biomarkers that can support earlier detection and risk stratification. This exploratory study investigated the combined diagnostic performance of circulating microRNA-155 (miR-155) and vascular endothelial growth factor (VEGF) in CRC, with mechanistic support from molecular docking and integration into an AI-based predictive model. In a case-control design, plasma levels of miR-155 and VEGF were quantified in CRC patients, individuals with benign colorectal conditions, and healthy controls. Diagnostic accuracy was assessed using ROC curve analysis, with comparisons across subgroup analyses (CRC vs benign, CRC vs controls, CRC vs benign + controls). Molecular docking provided comparative predictions of miR-155 interactions with regulatory proteins (IL-13RA1, SOCS1, PTEN, BCL-6, TP53INP1). An AI model (logistic regression with L2 regularization, stratified tenfold cross-validation) integrated biomarkers with clinical factors to evaluate predictive performance. Both miR-155 and VEGF were significantly elevated in CRC patients compared with benign and control groups. Individually, miR-155 achieved an AUC of 0.85 and VEGF an AUC of 0.79; combined analysis improved performance (AUC = 0.93). Subgroup ROC analyses confirmed robust discriminatory power across clinically relevant comparisons. The AI-integrated model achieved the highest accuracy (AUC = 0.96) under cross-validation. Docking suggested preferential interactions of miR-155 with IL-13RA1, SOCS1, and PTEN, supporting their mechanistic involvement. miR-155 and VEGF show promise as synergistic biomarkers for CRC detection, particularly when integrated with clinical risk factors. Molecular docking provides hypothesis-generating mechanistic insights, while AI modeling demonstrates the potential of multi-parametric integration. Given the modest, single-center sample size and lack of external validation, these findings should be considered exploratory. Larger, multi-center validation studies are essential before clinical translation.

Keywords: Biomarkers; CA19-9; CEA; Colorectal cancer; Diagnostic performance; ELISA; Early detection; MiR-155; QRT-PCR; VEGF.

Fig. 1

Receiver operating characteristic (ROC) curve analyses of miR-155, VEGF, and their combination across different subgroup comparisons. A CRC vs controls, B CRC vs benign colorectal conditions, and C CRC vs combined benign + control groups. In each panel, ROC curves for miR-155 (blue), VEGF (green), and the combined biomarker model (red) are shown. The combination consistently achieved the highest diagnostic accuracy, with AUC values of 0.94 (CRC vs controls), 0.88 (CRC vs benign), and 0.92 (CRC vs combined benign + controls). Individual biomarker performance was also strong: miR-155 yielded AUC values of 0.89, 0.81, and 0.85, while VEGF yielded 0.84, 0.77, and 0.79, respectively. These findings highlight the superior discriminatory power of the combined biomarker panel across clinically relevant comparisons

Fig. 2

Performance of the AI-based predictive model for CRC detection. A Bar chart showing the diagnostic performance metrics of the final AI model (logistic regression with L2 regularization): accuracy 92.3%, sensitivity 93.5%, specificity 91.2%, precision 91.0%, F1-score 92.2%, and ROC AUC 0.96. B Receiver operating characteristic (ROC) curve comparison of individual biomarkers (miR-155, VEGF), their combination, and the AI-integrated model. The AI model demonstrated the highest diagnostic accuracy (AUC = 0.96), outperforming miR-155 alone (AUC = 0.85), VEGF alone (AUC = 0.79), and the combined biomarker model (AUC = 0.93)

Fig. 3

Molecular docking analysis of miR-155 interactions with key regulatory proteins involved in colorectal cancer progression. Predicted binding affinities are shown as relative HDOCK scores. Among the tested proteins, IL-13RA1 exhibited the strongest interaction, followed by SOCS1 and PTEN, while BCL-6 and TP53INP1 demonstrated comparatively weaker but still favorable binding. These results highlight potential mechanistic pathways by which miR-155 may influence VEGF-driven angiogenesis, immune modulation, and tumor progression. Scores are interpreted comparatively, with more negative values reflecting stronger predicted interactions. For guidance, scores between –200 and –275 are considered moderate, while scores below –275 are considered promising

Fig. 3

Molecular docking analysis of miR-155 interactions with key regulatory proteins involved in colorectal cancer progression. Predicted binding affinities are shown as relative HDOCK scores. Among the tested proteins, IL-13RA1 exhibited the strongest interaction, followed by SOCS1 and PTEN, while BCL-6 and TP53INP1 demonstrated comparatively weaker but still favorable binding. These results highlight potential mechanistic pathways by which miR-155 may influence VEGF-driven angiogenesis, immune modulation, and tumor progression. Scores are interpreted comparatively, with more negative values reflecting stronger predicted interactions. For guidance, scores between –200 and –275 are considered moderate, while scores below –275 are considered promising