Liu Tingting, Zhong Lifan, Sun Xizhe, He Zhijiang, Lv Witiao, Deng Liyun, Chen Yanfei
Hainan Pharmaceutical Research and Development Science and Technology Park, Hainan Medical University, Haikou, Hainan, 571199, China.
Department of Orthopedics, Hainan Provincial Corps Hospital of Chinese People's Armed Police Force, Haikou, Hainan, 570203, China.
NPJ Precis Oncol. 2025 Aug 22;9(1):297. doi: 10.1038/s41698-025-01058-6.
Computational oncology advances multi-targeted therapies for Colon Cancer (CC) by leveraging molecular data and identifying potential drug candidates. However, challenges persist in understanding CC molecular pathways and identifying essential genes. This research integrates biomarker signatures from high-dimensional gene expression, mutation data, and protein interaction networks. The research study employs Adaptive Bacterial Foraging (ABF) optimization to refine search parameters, maximizing the predictive accuracy of therapeutic outcomes. The CatBoost algorithm efficiently classifies patients based on molecular profiles and predicts drug responses. The ABF-CatBoost integration facilitates a multi-targeted therapeutic approach, addressing drug resistance by analyzing mutation patterns, adaptive resistance mechanisms, and conserved binding sites. External validation datasets assess predictive accuracy and generalizability. The results demonstrated that the proposed system outperformed traditional Machine Learning models, such as Support Vector Machine and Random Forest, in terms of accuracy (98.6%), specificity (0.984), sensitivity (0.979), and F1-score (0.978). The model predicts toxicity risks, metabolism pathways, and drug efficacy profiles, ensuring safer and more effective treatments. The artificial intelligence model personalizes therapy by leveraging patient-specific molecular profiles, optimizing drug selection and dosage while minimizing side effects. By altering the biomarker selection and pathway analysis components, this computational framework is modified for other cancers, expanding its application and impact in personalized cancer treatment. It also improves precision medicine in CC therapy, speeding up drug discovery and improving therapeutic outcomes.
计算肿瘤学通过利用分子数据和识别潜在的候选药物，推进了结肠癌（CC）的多靶点治疗。然而，在理解CC分子途径和识别关键基因方面仍然存在挑战。本研究整合了来自高维基因表达、突变数据和蛋白质相互作用网络的生物标志物特征。该研究采用自适应细菌觅食（ABF）优化来优化搜索参数，最大限度地提高治疗结果的预测准确性。CatBoost算法根据分子特征有效地对患者进行分类，并预测药物反应。ABF-CatBoost整合促进了多靶点治疗方法，通过分析突变模式、适应性耐药机制和保守结合位点来解决耐药问题。外部验证数据集评估预测准确性和通用性。结果表明，所提出的系统在准确性（98.6%）、特异性（0.984）、敏感性（0.979）和F1分数（0.978）方面优于传统机器学习模型，如支持向量机和随机森林。该模型预测毒性风险、代谢途径和药物疗效概况，确保更安全、更有效的治疗。人工智能模型通过利用患者特异性分子特征实现个性化治疗，优化药物选择和剂量，同时将副作用降至最低。通过改变生物标志物选择和途径分析组件，该计算框架针对其他癌症进行了修改，扩大了其在个性化癌症治疗中的应用和影响。它还提高了CC治疗中的精准医学水平，加速了药物发现并改善了治疗结果。
