Exploration and validation of the Ki67, Her-2, and mutant P53 protein-based risk model, nomogram and lymph node metastasis model for predicting colorectal cancer progression and prognosis.

Introductions: Identifying biological markers of colorectal cancer (CRC) development and prognosis and exploring the intrinsic connection between these molecular markers and CRC progression is underway. However, a single molecular tumor marker is often difficult to assess and predict the progression and prognosis of CRC. Consequently, a combination of tumor-related markers is much needed. Ki67, Her-2, and mutant P53 (MutP53) proteins play pivotal roles in CRC occurrence, progression and prognosis. Methods: Based on the expressions by immunochemistry, we developed a risk model, nomogram and lymph node metastasis model by R software and Pythons to explore the value of these proteins in predicting CRC progression, prognosis, and examined the relationship of these proteins with the CRC clinicopathological features from 755 (training set) and 211 CRC (validation set) patients collected from the hospital. Results: We found that Ki67 expression was significantly correlated with T-stage, N-stage, TNM-stage, vascular invasion, organization differentiation, and adenoma carcinogenesis. Moreover, Her-2 expression was significantly correlated with T-stage, N-stage, TNM-stage, vascular and nerve invasion, pMMR/dMMR, signet ring cell carcinoma, and organization differentiation. MutP53 expression was significantly correlated with T-stage, N-stage, TNM-stage, vascular and nerve invasion, adenoma carcinogenesis, signet ring cell carcinoma, organization differentiation, and pMMR/dMMR. Increased expression of each of the protein indicated a poor prognosis. The established risk model based on the three key proteins showed high predictive value for determining the pathological characteristics and prognosis of CRC and was an independent influencer for prognosis. The nomogram prediction model, which was based on the risk model, after sufficient evaluation, showed more premium clinical value for predicting prognosis. Independent cohort of 211 CRC patients screened from the hospital verified the strong predictive efficacy of these models. The utilization of the XGBoost algorithm in a lymph node metastasis model, which incorporates three crucial proteins, demonstrated a robust predictive capacity for lymph node metastasis. Discussion: The risk model, nomogram and lymph node metastasis model have all provided valuable insights into the involvement of these three key proteins in the progression and prognosis of CRC. Our study provides a theoretical basis for further screening of effective models that utilize biological markers of CRC.