Gupta Tushar, Sharma Priyanka, Malik Sheeba, Pant Pradeep
Department of Biotechnology, Bennett University, Greater Noida 201310, Uttar Pradesh, India.
Department of Biochemistry and Cellular and Molecular Biology, The University of Tennessee, Knoxville, Tennessee 37996, United States.
Mol Pharm. 2025 Jul 7;22(7):4076-4090. doi: 10.1021/acs.molpharmaceut.5c00343. Epub 2025 Jun 18.
Aptamers are short, single-stranded DNA or RNA molecules known for their high specificity and affinity toward target biomolecules, making them powerful tools in drug discovery, diagnostics, and biosensing. However, conventional aptamer selection methods such as SELEX (Systematic Evolution of Ligands by EXponential Enrichment) are often labor-intensive, time-consuming, and resource-demanding. To overcome these limitations, we introduce a novel AI-driven aptamer optimization pipeline (AIoptamer: AI-driven optimization of aptamers) that integrates artificial intelligence with advanced classical computational approaches to accelerate aptamer discovery and design. The workflow begins with a known aptamer-host complex and systematically generates all possible aptamer sequence variants to target the same host. These variants are then screened using AI-based models that rank them based on sequence features and predicted binding affinity. Top candidates undergo structural modeling through CHIMERA_NA, an in-house mutagenesis tool designed to perform structural mutations in nucleic acids. The modeled structures are further evaluated using PredPRBA, a deep learning-based scoring function tailored for RNA-protein binding affinity prediction and PDA-Pred, a machine learning based model for predicting DNA-protein binding affinity. The highest-ranking aptamer-host complexes are then refined through molecular dynamics (MD) simulations to assess structural stability and interaction strength at the atomic level. Our pipeline demonstrates effectiveness across both RNA and DNA aptamer complexes, offering a generalized and robust framework for aptamer optimization. By combining AI-powered prediction with conventional computational techniques, our method advances the rational design of aptamers and significantly reduces reliance on traditional experimental trial-and-error strategies, making aptamer optimization faster, scalable and more efficient.
适体是短的单链DNA或RNA分子，以其对靶标生物分子的高特异性和亲和力而闻名，使其成为药物发现、诊断和生物传感领域的强大工具。然而，传统的适体筛选方法，如SELEX（指数富集配体系统进化），通常劳动强度大、耗时且资源需求高。为了克服这些局限性，我们引入了一种新型的人工智能驱动的适体优化流程（AIoptamer：人工智能驱动的适体优化），该流程将人工智能与先进的经典计算方法相结合，以加速适体的发现和设计。该工作流程从已知的适体-宿主复合物开始，系统地生成所有可能靶向同一宿主的适体序列变体。然后使用基于人工智能的模型对这些变体进行筛选，该模型根据序列特征和预测的结合亲和力对它们进行排名。顶级候选物通过CHIMERA_NA进行结构建模，CHIMERA_NA是一种内部诱变工具，旨在对核酸进行结构突变。使用PredPRBA（一种为RNA-蛋白质结合亲和力预测量身定制的基于深度学习的评分函数）和PDA-Pred（一种用于预测DNA-蛋白质结合亲和力的基于机器学习的模型）对建模结构进行进一步评估。然后通过分子动力学（MD）模拟对排名最高的适体-宿主复合物进行优化，以评估原子水平上的结构稳定性和相互作用强度。我们的流程在RNA和DNA适体复合物中均显示出有效性，为适体优化提供了一个通用且强大的框架。通过将人工智能驱动的预测与传统计算技术相结合，我们的方法推进了适体的合理设计，并显著减少了对传统实验试错策略的依赖，使适体优化更快、可扩展且更高效。
