机器学习在化学合成及表征中的应用

doi:10.16183/j.cnki.jsjtu.2023.078

摘要/Abstract

摘要：

自动化化学合成是化学领域长期追求的目标之一.近年来,机器学习的出现使得实现这一目标有了可能.以数据驱动为核心的机器学习借助计算机学习海量化学数据中的信息,寻找信息之间的客观联系和规律,根据已有规律和信息训练生成模型,借助模型预测分析需解决的实际问题.机器学习因其出色的计算预测能力,帮助化学工作者快速高效解决化学合成问题,加快研究进程.机器学习的出现和发展对化学合成及表征领域展示出强大的研究助力作用,但目前并不存在通用性极强的机器学习模型,化学工作者仍需根据实际情况选择不同模型进行训练学习.从监督学习、无监督学习、半监督学习、强化学习等机器学习的角度,向化学工作者展示常见学习方法在化学合成及表征中应用的最佳案例,帮助其利用机器学习知识进一步拓宽研究思路.

关键词: 机器学习, 化学合成, 监督学习, 强化学习

Abstract:

Automated chemical synthesis is one of the long-term goals pursued in the field of chemistry. In recent years, the advent of machine learning (ML) has made it possible to achieve this goal. Data-driven ML uses computers to learn relative information in massive chemical data, find objective connections between information, train models by using objective connections, and analyze the actual problems which can be solved according to these models. With its excellent computational prediction capabilities, ML helps chemists solve chemical synthesis problems quickly and efficiently and accelerate the research process. The emergence and development of ML has shown a strong research assistance in the field of chemical synthesis and characterization. However, there is no highly versatile ML model at present, and chemists still need to choose different models for training and learning according to actual situations. This paper aims to show chemists the best cases of common learning methods in chemical synthesis and characterization from the perspective of ML, such as supervised learning, unsupervised learning, semi-supervised learning, reinforcement learning, etc., and help them use ML knowledge to further broaden their research ideas.

Key words: machine learning (ML), chemistry synthesis, supervised learning, reinforcement learning

中图分类号:

O621
TP 3-05

孙婕, 李子昊, 张书宇. 机器学习在化学合成及表征中的应用[J]. 上海交通大学学报, 2023, 57(10): 1231-1244.

SUN Jie, LI Zihao, ZHANG Shuyu. Application of Machine Learning in Chemical Synthesis and Characterization[J]. Journal of Shanghai Jiao Tong University, 2023, 57(10): 1231-1244.

图/表 9

图1

表1

图2

图3

图4

图5

图6

图7

表2

ML在化学合成及表征领域的应用

ML算法	注意事项/适用范围	应用实例
SL-回归	研究自变量和其他变量之间的关系.一般使用不同模型进行拟合和交叉验证获得最优模型,具有很强的鲁棒性和容错性.需要考虑变量之间的相关性时采用多层回归,常用的NN模型能够无限逼近复杂的非线性模型,并行处理能力强,但是需要大量数据,输出结果的可解释性较弱.	逻辑回归:预测催化反应产率^[73] 多元LR:预测不对称反应中的对映选择性的关键参数^[74?-76] 多种回归模型进行比较:预测反应产率^[77] NN:正向反应预测^[78] NN:从反应底物预测产物^[5] NN:预测有机分子亲核性^[79]
SL-分类	对待测数据进行分类,通常是几种算法之间比较评估得出最优模型用于后续预测分析.RF算法可以保证分类节点特征的最优性但要避免过拟合现象,ET算法可以使节点特征选择具有随机性和最优性.SVM算法可以处理非线性数据,但需要进行线性化操作,将其转化为高维线性数据.	RF、SVM、ANN:预测有机分子的水溶性^[80] RF、SVM:预测交叉偶联反应各种潜在抑制配体的反应性能^[39] RF:有机化合物的紫外-可见光谱分类^[81] RF:设计催化剂^[82] RF:预测反应类型^[83] ET、RF:不对称区域选择性预测^[40]
贝叶斯理论	以贝叶斯公式为核心,模型容易理解,对小规模数据表现良好,过程简单,适合于小规模数据集的多分类问题,需要注意使用该理论时要有独立分布的假设前提.	贝叶斯优化器:优化反应条件^[84] 贝叶斯图卷积网络:预测分子表皮生长因子受体抑制活性^[85] 贝叶斯学习:预测金属位点反应性质^[86]
RL	从环境中学习信息,比其他方法更加智能化.奖励函数的设计是整个过程的核心,适合用于反应条件的优化问题,但采样数据的效率不高.	RL:迭代化学反应结果,优化化学反应^[87] 分子图+RL:分子设计^[88] SMILES字符串+RL:药物设计合成^[89] 分子图+RL:药物设计合成^[90]

表2

参考文献 90

[1]	JORDAN M I, MITCHELL T M. Machine learning: Trends, perspectives, and prospects[J]. Science, 2015, 349(6245): 255-260. doi: 10.1126/science.aaa8415 pmid: 26185243
[2]	BUTLER K T, DAVIES D W, CARTWRIGHT H, et al. Machine learning for molecular and materials science[J]. Nature, 2018, 559(7715): 547-555. doi: 10.1038/s41586-018-0337-2
[3]	WANG X R, LI Y Q, QIU J Z, et al. RetroPrime: A Diverse, plausible and Transformer-based method for Single-Step retrosynthesis predictions[J]. Chemical Engineering Journal, 2021, 420: 129845. doi: 10.1016/j.cej.2021.129845 URL
[4]	SCHWALLER P, LAINO T, GAUDIN T, et al. Molecular transformer: A model for uncertainty-calibrated chemical reaction prediction[J]. ACS Central Science, 2019, 5(9): 1572-1583. doi: 10.1021/acscentsci.9b00576 pmid: 31572784
[5]	WEI J N, DUVENAUD D, ASPURU-GUZIK A. Neural networks for the prediction of organic chemistry reactions[J]. ACS Central Science, 2016, 2(10): 725-732. pmid: 27800555
[6]	NAM J, KIM J. Linking the neural machine translation and the prediction of organic chemistry reactions[EB/OL].(2016-12-29) [2022-03-05].https://arxiv.org/abs/1612.09529.
[7]	LIU B W, RAMSUNDAR B, KAWTHEKAR P, et al. Retrosynthetic reaction prediction using neural sequence-to-sequence models[J]. ACS Central Science, 2017, 3(10): 1103-1113. doi: 10.1021/acscentsci.7b00303 pmid: 29104927
[8]	SEGLER M, PREUß M, WALLER M P. Towards "AlphaChem": Chemical synthesis planning with tree search and deep neural network policies[EB/OL]. (2017-01-31)[2022-03-05]. https://arxiv.org/abs/1702.00020.
[9]	SANCHEZ-LENGELING B, OUTEIRAL C, GUIMARAES G L, et al. Optimizing distributions over molecular space. An Objective-Reinforced Generative Adversarial Network for Inverse-design Chemistry (ORGANIC)[EB/OL]. (2017-08-18)[2022-03-05]. https://chemrxiv.org/engage/chemrxiv/article-details/60c73d91702a9beea7189bc2.
[10]	SARKER I H, HOQUE M M, UDDIN M K, et al. Mobile data science and intelligent apps: Concepts, AI-based modeling and research directions[J]. Mobile Networks and Applications, 2021, 26(1): 285-303. doi: 10.1007/s11036-020-01650-z
[11]	SARKER I H, KAYES A S M, BADSHA S, et al. Cybersecurity data science: An overview from machine learning perspective[J]. Journal of Big Data, 2020, 7(1): 41. doi: 10.1186/s40537-020-00318-5
[12]	WARR W A. A short review of chemical reaction database systems, computer-aided synthesis design, reaction prediction and synthetic feasibility[J]. Molecular Informatics, 2014, 33(6/7): 469-476. doi: 10.1002/minf.v33.6/7 URL
[13]	WEININGER D. SMILES, a chemical language and information system. 1. introduction to methodology and encoding rules[J]. Journal of Chemical Information & Computer Sciences, 1988, 28(1): 31-36.
[14]	JELIAZKOVA N, KOCHEV N. AMBIT-SMARTS: Efficient searching of chemical structures and fragments[J]. Molecular Informatics, 2011, 30(8): 707-720. doi: 10.1002/minf.201100028 pmid: 27467262
[15]	LI Z H, LI Q Z, BAI H Y, et al. Synthetic strategies and mechanistic studies of axially chiral styrenes[J]. Chem Catalysis, 2023, 3: 100594. doi: 10.1016/j.checat.2023.100594 URL
[16]	TOMBERG A, JOHANSSON M J, NORRBY P O. A predictive tool for electrophilic aromatic substitutions using machine learning[J]. The Journal of Organic Chemistry, 2019, 84(8): 4695-4703. doi: 10.1021/acs.joc.8b02270 URL
[17]	LI Q Z, LI Z H, KANG J C, et al. Ni-catalyzed, enantioselective three-component radical relayed reductive coupling of alkynes: Synthesis of axially chiral styrenes[J]. Chem Catalysis, 2022, 2(11): 3185-3195. doi: 10.1016/j.checat.2022.09.020 URL
[18]	CHEN C, ZHANG S Y, LI S X, et al. Electroreductive fluoroalkylative heteroarylation of unactivated alkenes via an unconventional remote heteroaryl migration[J]. Cell Reports Physical Science, 2023, 4(5): 101385. doi: 10.1016/j.xcrp.2023.101385 URL
[19]	SARKER I H. Machine learning: Algorithms, real-world applications and research directions[J]. SN Computer Science, 2021, 2(3): 160. doi: 10.1007/s42979-021-00592-x
[20]	HAN J, PEI J, TONG H. Data mining: Concepts and techniques[M]. San Mateo: Morgan Kaufmann, 2022.
[21]	WANG H F, HU D J. Comparison of SVM and LS-SVM for regression[C]// 2005 International Conference on Neural Networks and Brain. Beijing, China: IEEE, 2005: 279-283.
[22]	CHERKASSKY V, MA Y Q. Practical selection of SVM parameters and noise estimation for SVM regression[J]. Neural Networks, 2004, 17(1): 113-126. pmid: 14690712
[23]	CORTES C, VAPNIK V. Support-vector networks[J]. Machine Learning, 1995, 20(3): 273-297.
[24]	HAMERLY G, ELKAN C. Learning the k in k-means[C]// Proceedings of the 16th International Conference on Neural Information Processing Systems. Cambridge, MA, USA: MIT Press, 2003: 281-288.
[25]	MONTGOMERY D C, PECK E A, VINING G G. Introduction to linear regression analysis[M]. 6th ed. Hoboken: Wiley, 2021.
[26]	MAHALAXMI K V K, REKHA K S. Comparison of logistic regression and artificial neural network for modelling credit card data set with the identification of precise fraudulent[C]//2022 International Conference on Business Analytics for Technology and Security. Dubai: IEEE, 2022: 1-5.
[27]	MADHULATHA T S. An overview on clustering methods[EB/OL].(2012-05-05)[2022-03-05]. https://arxiv.org/abs/1205.1117.
[28]	LAPAN M. Deep reinforcement learning hands-on:Apply modern RL methods, with deep Q-networks, value iteration, policy gradients, TRPO, AlphaGo Zero and more[M]. Mumbai: Packt Publishing, 2018.
[29]	SILVER D, HUANG A, MADDISON C J, et al. Mastering the game of Go with deep neural networks and tree search[J]. Nature, 2016, 529(7587): 484-489. doi: 10.1038/nature16961
[30]	KALASHNIKOV D, IRPAN A, PASTOR P, et al. Scalable deep reinforcement learning for vision-based robotic manipulation[EB/OL]. (2018-06-27)[2022-03-05].https://arxiv.org/abs/1806.10293v3.
[31]	COREY E J, WIPKE W T. Computer-assisted design of complex organic syntheses[J]. Science, 1969, 166(3902): 178-192. pmid: 17731475
[32]	SADYBEKOV A V, KATRITCH V. Computational approaches streamlining drug discovery[J]. Nature, 2023, 616(7958): 673-685. doi: 10.1038/s41586-023-05905-z
[33]	刘伊迪, 杨骐, 李遥, 等. 机器学习在有机化学中的应用[J]. 有机化学, 2020, 40(11): 3812-3827. doi: 10.6023/cjoc202006051
	LIU Yidi, YANG Qi, LI Yao, et al. Application of machine learning in organic chemistry[J]. Chinese Journal of Organic Chemistry, 2020, 40(11): 3812-3827. doi: 10.6023/cjoc202006051
[34]	BREIMAN L. Random forests[J]. Machine Learning, 2001, 45(1): 5-32. doi: 10.1023/A:1010933404324 URL
[35]	SINGH S, PAREEK M, CHANGOTRA A, et al. A unified machine-learning protocol for asymmetric catalysis as a proof of concept demonstration using asymmetric hydrogenation[J]. Proceedings of the National Academy of Sciences of the United States of America, 2020, 117(3): 1339-1345. doi: 10.1073/pnas.1916392117 pmid: 31915295
[36]	PEDREGOSA F, VAROQUAUX G, GRAMFORT A, et al. Scikit-learn: Machine learning in python[EB/OL]. (2012-01-02)[2022-03-05].https://arxiv.org/abs/1201.0490.
[37]	KANG B, SEOK C, LEE J Y. Prediction of molecular electronic transitions using random forests[J]. Journal of Chemical Information and Modeling, 2020, 60(12): 5984-5994. doi: 10.1021/acs.jcim.0c00698 pmid: 33090804
[38]	LI X, ZHANG S Q, XU L C, et al. Predicting regioselectivity in radical C—H functionalization of heterocycles through machine learning[J]. Angewandte Chemie International Edition, 2020, 59(32): 13253-13259. doi: 10.1002/anie.v59.32 URL
[39]	AHNEMAN D T, ESTRADA J G, LIN S S, et al. Predicting reaction performance in C—N cross-coupling using machine learning[J]. Science, 2018, 360(6385): 186-190. doi: 10.1126/science.aar5169 URL
[40]	XU L C, FREY J, HOU X Y, et al. Enantioselectivity prediction of pallada-electrocatalysed C—H activation using transition state knowledge in machine learning[J]. Nature Synthesis, 2023, 2(4): 321-330. doi: 10.1038/s44160-022-00233-y
[41]	LECUN Y, BOSER B, DENKER J, et al. Handwritten digit recognition with a back-propagation network[C]//Proceedings of the 2nd International Conference on Neural Information Processing Systems. Cambridge, MA, USA: MIT Press, 1989: 396-404.
[42]	LECUN Y, BOTTOU L, ORR G B, et al. Efficient BackProp[M]//Neural networks:Tricks of the trade. Berlin, Heidelberg: Springer, 2012: 9-48.
[43]	NAIR V, HINTON G E. Rectified linear units improve restricted boltzmann machines[C]//Proceedings of the 27th International Conference on International Conference on Machine Learning. Madison, WI, USA: Omnipress, 2010: 807-814.
[44]	WANG T, WU D J, COATES A, et al. End-to-end text recognition with convolutional neural networks[C]// Proceedings of the 21 st International Conference on Pattern Recognition. Tsukuba, Japan: IEEE, 2012: 3304-3308.
[45]	GU J X, WANG Z H, KUEN J, et al. Recent advances in convolutional neural networks[J]. Pattern Recognition, 2018, 77: 354-377. doi: 10.1016/j.patcog.2017.10.013 URL
[46]	ZINKEVICH M A, WEIMER M, SMOLA A, et al. Parallelized stochastic gradient descent[C]// Proceedings of the 23rd International Conference on Neural Information Processing Systems. Red Hook, NY, USA: Curran Associates Inc., 2010: 2595-2603.
[47]	HIROHARA M, SAITO Y, KODA Y, et al. Convolutional neural network based on SMILES representation of compounds for detecting chemical motif[J]. BMC Bioinformatics, 2018, 19(Sup.19): 526. doi: 10.1186/s12859-018-2523-5
[48]	WALLACH I, DZAMBA M, HEIFETS A. AtomNet: A deep convolutional neural network for bioactivity prediction in structure-based drug discovery[EB/OL].(2015-10-10)[2022-03-05].https://arxiv.org/abs/1510.02855.
[49]	HUGHES T B, MILLER G P, SWAMIDASS S J. Modeling epoxidation of drug-like molecules with a deep machine learning network[J]. ACS Central Science, 2015, 1(4): 168-180. doi: 10.1021/acscentsci.5b00131 pmid: 27162970
[50]	HUGHES T B, MILLER G P, SWAMIDASS S J. Site of reactivity models predict molecular reactivity of diverse chemicals with glutathione[J]. Chemical Research in Toxicology, 2015, 28(4): 797-809. doi: 10.1021/acs.chemrestox.5b00017 pmid: 25742281
[51]	XING S P, YU H X, LIU M, et al. Recognizing contamination fragment ions in liquid chromatography-tandem mass spectrometry data[J]. Journal of the American Society for Mass Spectrometry, 2021, 32(9): 2296-2305. doi: 10.1021/jasms.0c00478 URL
[52]	ZHENG X X, YANG Z X, YANG C, et al. Fast acquisition of high-quality nuclear magnetic resonance pure shift spectroscopy via a deep neural network[J]. The Journal of Physical Chemistry Letters, 2022, 13(9): 2101-2106. doi: 10.1021/acs.jpclett.2c00100 URL
[53]	BRONSTEIN M M, BRUNA J, LECUN Y, et al. Geometric deep learning: Going beyond euclidean data[J]. IEEE Signal Processing Magazine, 2017, 34(4): 18-42.
[54]	YING R, BOURGEOIS D, YOU J X, et al. GNNExplainer: Generating explanations for graph neural networks[EB/OL]. (2019-03-10)[2022-03-05]. https://arxiv.org/abs/1903.03894.
[55]	LI Y J, TARLOW D, BROCKSCHMIDT M, et al. Gated graph sequence neural networks[EB/OL]. (2015-11-17)[2022-03-05].https://arxiv.org/abs/1511.05493.
[56]	KIPF T N, WELLING M. Semi-supervised classification with graph convolutional networks[EB/OL].(2016-09-09)[2022-03-05].https://arxiv.org/abs/1609.02907.
[57]	DUVENAUD D, MACLAURIN D, AGUILERA-IPARRAGUIRRE J, et al. Convolutional networks on graphs for learning molecular fingerprints[C]//Proceedings of the 28th International Conference on Neural Information Processing Systems-Volume 2. Montreal, Canada: ACM, 2015: 2224-2232.
[58]	COLEY C, JIN W G, ROGERS L, et al. A graph-convolutional neural network model for the prediction of chemical reactivity[J]. Chemical Science, 2019, 10(2): 370-377. doi: 10.1039/c8sc04228d pmid: 30746086
[59]	SAEBI M, NAN B, HERR J E, et al. Graph neural networks for predicting chemical reaction performance[EB/OL]. (2021-05-01)[2022-03-05]. https://www.researchgate.net/publication/351635791_Graph_Neural_Networks_for_Predicting_Chemical_Reaction_Performance.
[60]	MATER A C, COOTE M L. Deep learning in chemistry[J]. Journal of Chemical Information and Modeling, 2019, 59(6): 2545-2559. doi: 10.1021/acs.jcim.9b00266 pmid: 31194543
[61]	ROSZAK R, BEKER W, MOLGA K, et al. Rapid and accurate prediction of pK_a values of C—H acids using graph convolutional neural networks[J]. Journal of the American Chemical Society, 2019, 141(43): 17142-17149. doi: 10.1021/jacs.9b05895 URL
[62]	WEN M J, BLAU S M, SPOTTE-SMITH E W C, et al. BonDNet: A graph neural network for the prediction of bond dissociation energies for charged molecules[J]. Chemical Science, 2021, 12(5): 1858-1868. doi: 10.1039/D0SC05251E URL
[63]	GRAMBOW C A, PATTANAIK L, GREEN W H. Deep learning of activation energies[J]. The Journal of Physical Chemistry Letters, 2020, 11(8): 2992-2997. doi: 10.1021/acs.jpclett.0c00500 URL
[64]	ZAHRT A F, HENLE J J, ROSE B T, et al. Prediction of higher-selectivity catalysts by computer-driven workflow and machine learning[J]. Science, 2019, 363(6424): eaau5631. doi: 10.1126/science.aau5631 URL
[65]	VAN OTTERLO M, WIERING M. Reinforcement learning and Markov decision processes[M]//Adaptation, learning, and optimization. Berlin, Heidelberg: Springer, 2012: 3-42.
[66]	SEGLER M H S, PREUSS M, WALLER M P. Planning chemical syntheses with deep neural networks and symbolic AI[J]. Nature, 2018, 555(7698): 604-610. doi: 10.1038/nature25978 URL
[67]	LI S, DENG M, LEE J, et al. Imaging through glass diffusers using densely connected convolutional networks[J]. Optica, 2018, 5(7): 803. doi: 10.1364/OPTICA.5.000803 URL
[68]	WU Z Q, RAMSUNDAR B, FEINBERG E N, et al. MoleculeNet: A benchmark for molecular machine learning[J]. Chemical Science, 2017, 9(2): 513-530. doi: 10.1039/C7SC02664A URL
[69]	RUDDIGKEIT L, VAN DEURSEN R, BLUM L C, et al. Enumeration of 166 billion organic small molecules in the chemical universe database GDB-17[J]. Journal of Chemical Information and Modeling, 2012, 52(11): 2864-2875. doi: 10.1021/ci300415d pmid: 23088335
[70]	SCHÜTT K T, KESSEL P, GASTEGGER M, et al. SchNetPack: A deep learning toolbox for atomistic systems[J]. Journal of Chemical Theory and Computation, 2019, 15(1): 448-455. doi: 10.1021/acs.jctc.8b00908 pmid: 30481453
[71]	BENDER A, SCHNEIDER N, SEGLER M, et al. Evaluation guidelines for machine learning tools in the chemical sciences[J]. Nature Reviews Chemistry, 2022, 6(6): 428-442. doi: 10.1038/s41570-022-00391-9 pmid: 37117429
[72]	CICHOŃSKA A, RAVIKUMAR B, ALLAWAY R J, et al. Crowdsourced mapping of unexplored target space of kinase inhibitors[J]. Nature Communications, 2021, 12(1): 1-18. doi: 10.1038/s41467-020-20314-w
[73]	YADA A, NAGATA K, ANDO Y, et al. Machine learning approach for prediction of reaction yield with simulated catalyst parameters[J]. Chemistry Letters, 2018, 47(3): 284-287. doi: 10.1246/cl.171130 URL
[74]	ZHANG C, SANTIAGO C B, CRAWFORD J M, et al. Enantioselective dehydrogenative heck arylations of trisubstituted alkenes with indoles to construct quaternary stereocenters[J]. Journal of the American Chemical Society, 2015, 137(50): 15668-15671. doi: 10.1021/jacs.5b11335 pmid: 26624236
[75]	PARK Y, NIEMEYER Z L, YU J Q, et al. Quantifying structural effects of amino acid ligands in Pd(II)-catalyzed enantioselective C—H functionalization reactions[J]. Organometallics, 2018, 37(2): 203-210. doi: 10.1021/acs.organomet.7b00751 URL
[76]	GUO J Y, MINKO Y, SANTIAGO C B, et al. Developing comprehensive computational parameter sets to describe the performance of pyridine-oxazoline and related ligands[J]. ACS Catalysis, 2017, 7(6): 4144-4151. doi: 10.1021/acscatal.7b00739 URL
[77]	EBI T, SEN A, DHITAL R N, et al. Design of experimental conditions with machine learning for collaborative organic synthesis reactions using transition-metal catalysts[J]. ACS Omega, 2021, 6(41): 27578-27586. doi: 10.1021/acsomega.1c04826 pmid: 34693179
[78]	COLEY C W, BARZILAY R, JAAKKOLA T S, et al. Prediction of organic reaction outcomes using machine learning[J]. ACS Central Science, 2017, 3(5): 434-443. doi: 10.1021/acscentsci.7b00064 pmid: 28573205
[79]	SAINI V, SHARMA A, NIVATIA D. A machine learning approach for predicting the nucleophilicity of organic molecules[J]. Physical Chemistry Chemical Physics, 2022, 24(3): 1821-1829. doi: 10.1039/d1cp05072a pmid: 34986215
[80]	PALMER D S, O’BOYLE N M, GLEN R C, et al. Random forest models to predict aqueous solubility[J]. Journal of Chemical Information and Modeling, 2007, 47(1): 150-158. doi: 10.1021/ci060164k pmid: 17238260
[81]	MAMEDE R, PEREIRA F, AIRES-DE-SOUSA J. Machine learning prediction of UV-Vis spectra features of organic compounds related to photoreactive potential[J]. Scientific Reports, 2021, 11(1): 1-11. doi: 10.1038/s41598-020-79139-8
[82]	BANERJEE S, SREENITHYA A, SUNOJ R B. Machine learning for predicting product distributions in catalytic regioselective reactions[J]. Physical Chemistry Chemical Physics, 2018, 20(27): 18311-18318. doi: 10.1039/c8cp03141j pmid: 29967920
[83]	ZHANG Q Y, AIRES-DE-SOUSA J. Structure-based classification of chemical reactions without assignment of reaction centers[J]. Journal of Chemical Information and Modeling, 2005, 45(6): 1775-1783. doi: 10.1021/ci0502707 URL
[84]	SHIELDS B J, STEVENS J, LI J, et al. Bayesian reaction optimization as a tool for chemical synthesis[J]. Nature, 2021, 590(7844): 89-96. doi: 10.1038/s41586-021-03213-y
[85]	RYU S, KWON Y, KIM W Y. A Bayesian graph convolutional network for reliable prediction of molecular properties with uncertainty quantification[J]. Chemical Science, 2019, 10(36): 8438-8446. doi: 10.1039/c9sc01992h pmid: 31803423
[86]	WANG S W, PILLAI H S, XIN H L. Bayesian learning of chemisorption for bridging the complexity of electronic descriptors[J]. Nature Communications, 2020, 11(1): 1-7. doi: 10.1038/s41467-019-13993-7
[87]	ZHOU Z P, LI X C, ZARE R N. Optimizing chemical reactions with deep reinforcement learning[J]. ACS Central Science, 2017, 3(12): 1337-1344. doi: 10.1021/acscentsci.7b00492 pmid: 29296675
[88]	SIMM G N C, PINSLER R, HERNÁNDEZ-LOBATO J M. Reinforcement learning for molecular design guided by quantum mechanics[EB/OL]. (2020-02-18)[2022-03-05]. https://arxiv.org/abs/2002.07717.
[89]	POPOVA M, ISAYEV O, TROPSHA A. Deep reinforcement learning forde novo drug design[J]. Science Advances, 2018, 4(7): eaap7885. doi: 10.1126/sciadv.aap7885 URL
[90]	ATANCE S R, DIEZ J V, ENGKVIST O, et al. De novo drug design using reinforcement learning with graph-based deep generative models[J]. Journal of Chemical Information and Modeling, 2022, 62(20): 4863-4872. doi: 10.1021/acs.jcim.2c00838 pmid: 36219571

描述符名称	表现形式	优势	不足	适用范围
SMILES^[13]字符 SMARTS^[14]字符 Inchl字符	SMILES: CC(OC1=CC=CC=C1)=O SMATRS: [C]-[C](-[O]-[C]1: [C]: [C]: [C]: [C]: [C]: 1)=[O] Inchl: 1S/C8H8O2/c1-7(9)10-8-5-3-2-4-6-8/h2-6H, 1H3	采用线性方法对分子进行表示,简单易操作;不同分子的SMILES不同,具有唯一性;占用内存小,节省存储空间.	丢失分子的三维信息;每个SMILES字符串对分子图的表示方法不唯一,即可从不同方向对分子图进行编码.	不需要分子空间信息;需要大量数据进行训练的模型.
分子指纹	图示①	采用比特量形式表示分子,编解码简单;能够表示分子的局部信息;分子的特征之间相互独立.	分子信息存在冗余,占用存储空间大;计算时间长,每次计算需要进行遍历.	擅长计算分子之间的相似性;描述分子的部分结构信息.
分子图	图示②	分子可视性强;描述符可解释性强;能够描述分子的三维信息.	信息传递更新过程慢,计算过程复杂.	图神经网络模型的输入;需要分子空间信息的场合.
量子化学描述符	过渡态能量^[15]、波函数、Fukui函数^[16]、分子键序、分子电荷等.	能够精准计算分子的化学和物理性质.	计算时间长;计算过程繁琐.	需精确描述分子性质的场合.