Assigning confidence to molecular property prediction.
Assigning confidence to molecular property prediction.
复制标题
赋予分子性质预测信心。
DOI：
10.1080/17460441.2021.1925247
复制
发表时间：
2021-09
影响因子：
6.3
通讯作者：
Aspuru-Guzik, Alan
中科院分区：
文献类型：
--
作者：
Nigam, AkshatKumar;Pollice, Robert;Hurley, Matthew F. D.;Hickman, Riley J.;Aldeghi, Matteo;Yoshikawa, Naruki;Chithrananda, Seyone;Voelz, Vincent A.;Aspuru-Guzik, Alan
关键词：
在过去的几十年里，计算建模得到了迅速发展，特别是在化学、材料科学和药物设计中预测分子特性。最近，机器学习技术已经成为一种强大且具有成本效益的策略，可以从现有的数据集中学习并对看不见的分子进行预测。因此，数据驱动技术的爆炸式增长提出了一个重要的问题：分子性质预测的可信度如何，以及什么技术可以用于此目的？在这项工作中，我们讨论了流行的策略，预测分子特性相关的药物设计，其相应的不确定性来源和方法来量化的不确定性和信心。首先，我们对评估置信度的考虑开始于数据集偏差和大小、数据驱动的属性预测和特征设计。接下来，我们将详细讨论通过分子对接进行的性质模拟和结合亲和力的自由能模拟。最后，我们研究了这些不确定性如何传播到生成模型，因为它们通常与属性预测因子相结合。在探索巨大的化学空间时，计算技术对于减少蛮力实验的高昂成本和时间至关重要。我们认为，评估属性预测模型中的不确定性是必不可少的，无论何时，依赖于高通量虚拟筛选的闭环药物设计活动部署。因此，考虑不确定性的来源会导致更好的实验验证，更可靠的预测和整个工作流程的更现实的期望。总体而言，这增加了对预测和设计的信心，并最终加速了药物设计。
Computational modeling has rapidly advanced over the last decades, especially to predict molecular properties for chemistry, material science and drug design. Recently, machine learning techniques have emerged as a powerful and cost-effective strategy to learn from existing datasets and perform predictions on unseen molecules. Accordingly, the explosive rise of data-driven techniques raises an important question: What confidence can be assigned to molecular property predictions and what techniques can be used for that purpose? In this work, we discuss popular strategies for predicting molecular properties relevant to drug design, their corresponding uncertainty sources and methods to quantify uncertainty and confidence. First, our considerations for assessing confidence begin with dataset bias and size, data-driven property prediction and feature design. Next, we discuss property simulation via molecular docking, and free-energy simulations of binding affinity in detail. Lastly, we investigate how these uncertainties propagate to generative models, as they are usually coupled with property predictors. Computational techniques are paramount to reduce the prohibitive cost and timing of brute-force experimentation when exploring the enormous chemical space. We believe that assessing uncertainty in property prediction models is essential whenever closed-loop drug design campaigns relying on high-throughput virtual screening are deployed. Accordingly, considering sources of uncertainty leads to better-informed experimental validations, more reliable predictions and to more realistic expectations of the entire workflow. Overall, this increases confidence in the predictions and designs and, ultimately, accelerates drug design.
登录
查看更多内容
影响因子：
4.1
作者：
BENNETT, CH
通讯作者：
BENNETT, CH
药物发现的迁移学习
影响因子：
7.3
作者：
Cai, Chenjing;Wang, Shiwei;Pei, Jianfeng
通讯作者：
Pei, Jianfeng
影响因子：
3.5
作者：
Bannan CC;Mobley DL;Skillman AG
通讯作者：
Skillman AG
影响因子：
18.2
作者：
Gómez-Bombarelli R;Wei JN;Duvenaud D;Hernández-Lobato JM;Sánchez-Lengeling B;Sheberla D;Aguilera-Iparraguirre J;Hirzel TD;Adams RP;Aspuru-Guzik A
通讯作者：
Aspuru-Guzik A
影响因子：
5.5
作者：
Beauchamp, Kyle A.;Lin, Yu-Shan;Das, Rhiju;Pande, Vijay S.
通讯作者：
Pande, Vijay S.