AI+Bio 25/3/5 文献速递|机器学习在小分子药物可开发性预测和超大规模药物库筛选中的作用综述

多客科技

作者：微信文章
关注AI+Biology线报，每日推送AI+Biology当日更新的最新文章和资讯，获取领域最新的技术进展
Deepmind推出Alphafold3官方教程，从原理到实操带你精通结构预测重磅｜Deepmind推出Alphafold3官方教程，从原理到实操带你精通结构预测数据决定成败：机器学习在小分子药物研发的未来｜Nature Computational Science最新综述
阿斯利康重磅｜抗体设计中的生成模型全面评测
一击即中！BindCraft 实现蛋白binder的one-shot设计（附完整protocol

---

1.PROPERMAB: an integrative framework for in silico prediction of antibody developability using machine learning


期刊：Mabs
链接：https://www.tandfonline.com/doi/full/10.1080/19420862.2025.2474521?af=R
总结：
该研究提出了一种计算框架 PROPERMAB，通过机器学习预测单克隆抗体的可开发性特性，如疏水相互作用色谱保留时间和高浓度黏度。研究表明，该方法可高效评估抗体的可开发性，并扩展到大规模抗体序列数据集，从而加速从发现到临床的开发进程。
摘要：
Selection of lead therapeutic molecules is often driven predominantly by pharmacological efficacy and safety. Candidate developability, such as biophysical properties that affect the formulation of the molecule into a product, is usually evaluated only toward the end of the drug development pipeline...

---

2.Integrating protein dynamics into structure-based drug design via full-atom stochastic flows


期刊：arxiv
链接：https://arxiv.org/abs/2503.03989
总结：
该研究提出了一种生成式建模方法 DynamicFlow，以全原子流模型结合分子动力学模拟，捕捉蛋白质结合口袋的构象变化，从而优化基于结构的药物设计（SBDD）。这一方法有助于发现新的配体分子，并为传统 SBDD 提供更优质的输入，提高药物开发的有效性。
摘要：
The dynamic nature of proteins, influenced by ligand interactions, is essential for comprehending protein function and progressing drug discovery. Traditional structure-based drug design (SBDD) approaches typically target binding sites with rigid structures, limiting their practical application in drug development...

---

3.AFsample2 predicts multiple conformations and ensembles with AlphaFold2


期刊：Communications Biology
链接：https://doi.org/10.1038/s42003-025-07791-9
总结：
AFsample2 通过随机掩蔽 MSA 列，减少共进化信号，从而提高 AlphaFold2 预测的构象多样性。该方法在多种蛋白质上取得了良好效果，提高了终态预测的质量，并增强了中间态的多样性，使其成为探索多构象蛋白质结构的重要工具。
摘要：
Understanding protein dynamics and conformational states is crucial for insights into biological processes and disease mechanisms, which can aid drug development. Recently, several methods have been devised to broaden the conformational predictions made by AlphaFold2 (AF2)...

---

4.UPDATE: the impact of conformer quality on learned representations of molecular conformer ensembles


期刊：arxiv
链接：https://arxiv.org/abs/2502.13220
总结：
该研究探讨了分子构象质量对机器学习模型预测 3D 结构相关性质的影响。研究结果表明，低质量构象对高质量构象的预测存在一定误差，但通过优化训练数据质量，可以提高模型在分子性质预测中的表现。
摘要：
Training machine learning models to predict properties of molecular conformer ensembles is an increasingly popular strategy to accelerate the conformational analysis of drug-like small molecules, reactive organic substrates, and homogeneous catalysts...

---

5.Progress of machine learning in the application of small molecule druggability prediction


期刊：European Journal of Medicinal Chemistry
链接：https://doi.org/10.1016/j.ejmech.2025.117269
总结：
该综述介绍了机器学习在小分子药物可开发性预测中的应用，包括溶解度、活性、毒性、药代动力学等方面的预测模型。此外，文章讨论了 ML 在超大规模药物库筛选中的作用，并展望了未来的发展趋势。
摘要：
Machine learning (ML) has become an important tool for predicting the pharmaceutical properties of small molecules. Recent advancements in ML algorithms enable the rapid and accurate evaluation of solubility, activity, toxicity, pharmacokinetics, and other molecular properties through ML-based models...

---

6.Benchmarking large language models on multiple tasks in bioinformatics NLP with prompting


期刊：arxiv
链接：https://arxiv.org/abs/2503.04013
总结：
该研究提出了一种基于提示工程的生物信息学大语言模型（LLM）基准测试框架 Bio-Benchmark，涵盖 30 项关键任务，并评估了 GPT-4o、Llama-3.1-70B 等主流 LLM 的表现。研究发现，LLM 在某些生物任务上表现优异，但仍需改进提示工程策略以优化性能。
摘要：
Large language models (LLMs) have become important tools in solving biological problems, offering improvements in accuracy and adaptability over conventional methods. Several benchmarks have been proposed to evaluate the performance of these LLMs...

---

7.Collaborative expert LLMs guided multi-objective molecular optimization


期刊：arxiv
链接：https://arxiv.org/abs/2503.03503
总结：
该研究提出 MultiMol，一个基于 LLM 的协作优化系统，用于多目标分子优化。MultiMol 结合数据驱动和文献指导的 LLM 代理，在六项任务上均显著优于现有方法，并在真实案例中优化了 XAC 选择性和 Saquinavir 的生物利用度，显示出巨大的药物开发潜力。
摘要：
Molecular optimization is a crucial yet complex and time-intensive process that often acts as a bottleneck for drug development. Traditional methods rely heavily on trial and error, making multi-objective optimization both time-consuming and resource-intensive...

---

8.TEDDY: a family of foundation models for understanding single cell biology


期刊：arxiv
链接：https://arxiv.org/abs/2503.03485
总结：
TEDDY 是一组基于 Transformer 的单细胞基础模型，利用 1.16 亿个细胞的超大规模数据集进行训练。研究表明，模型在疾病状态识别任务上表现出色，优于现有方法，并在细胞分类任务上也有一定改进，推动了单细胞数据分析的发展。
摘要：
Understanding the biological mechanism of disease is critical for medicine, and in particular drug discovery. AI-powered analysis of genome-scale biological data hold great potential in this regard...

---

9.Can frontier LLMs replace annotators in biomedical text mining? Analyzing challenges and exploring solutions


期刊：arxiv
链接：https://arxiv.org/abs/2503.03261
总结：
该研究探讨了 LLM 在生物医学文本挖掘中的作用，并分析了其在学习数据集特定细微差别、格式化要求和严格遵守标注指南方面的挑战。研究表明，通过优化提示工程和自动化标注流程，LLM 可部分替代人工标注，提高效率。
摘要：
Large language models (LLMs) can perform various natural language processing (NLP) tasks through in-context learning without relying on supervised data. However, multiple previous studies have reported suboptimal performance of LLMs in biological text mining...

---

10.Transformers for molecular property prediction: domain adaptation efficiently improves performance


期刊：arxiv
链接：https://arxiv.org/abs/2503.03360
总结：
该研究探讨了 Transformer 预训练数据规模对分子性质预测的影响，并提出了领域自适应策略，以小规模高相关数据集微调模型，从而在多个 ADME 端点上显著提升预测性能。
摘要：
Most of the current transformer-based chemical language models are pre-trained on millions to billions of molecules. However, the improvement from such scaling in dataset size is not confidently linked to improved molecular property prediction...



1.















DiffSBDD 是一种基于对称性扩散模型的新方法，通过 3D 条件生成问题，拓展了结构药物设计的适用性，为药物生成提供了更广泛的解决方案。