Machine learning-based solubility prediction and methodology evaluation of active pharmaceutical ingredients in industrial crystallization

doi:10.1007/s11705-021-2083-5

Front. Chem. Sci. Eng.

2022, Vol. 16

Issue (4) : 523-535 https://doi.org/10.1007/s11705-021-2083-5

RESEARCH ARTICLE

Machine learning-based solubility prediction and methodology evaluation of active pharmaceutical ingredients in industrial crystallization

Yiming Ma^1,², Zhenguo Gao^1,², Peng Shi^1,², Mingyang Chen^1,², Songgu Wu^1,², Chao Yang³, Jing-Kang Wang^1,², Jingcai Cheng³(

), Junbo Gong^1,²(

)

¹. School of Chemical Engineering and Technology, State Key Laboratory of Chemical Engineering, Tianjin University, Tianjin 300072, China
². The Co-Innovation Center of Chemistry and Chemical Engineering of Tianjin, Tianjin 300072, China
³. Key Laboratory of Green Process and Engineering, Institute of Process Engineering, Chinese Academy of Sciences, Beijing 100190, China

Download: PDF(1361 KB) HTML
Export: BibTeX | EndNote | Reference Manager | ProCite | RefWorks

Abstract

Solubility has been widely regarded as a fundamental property of small molecule drugs and drug candidates, as it has a profound impact on the crystallization process. Solubility prediction, as an alternative to experiments which can reduce waste and improve crystallization process efficiency, has attracted increasing attention. However, there are still many urgent challenges thus far. Herein we used seven descriptors based on understanding dissolution behavior to establish two solubility prediction models by machine learning algorithms. The solubility data of 120 active pharmaceutical ingredients (APIs) in ethanol were considered in the prediction models, which were constructed by random decision forests and artificial neural network with optimized data structure and model accuracy. Furthermore, a comparison with traditional prediction methods including the modified solubility equation and the quantitative structure-property relationships model was carried out. The highest accuracy shown by the testing set proves that the ML models have the best solubility prediction ability. Multiple linear regression and stepwise regression were used to further investigate the critical factor in determining solubility value. The results revealed that the API properties and the solute-solvent interaction both provide a nonnegligible contribution to the solubility value.

Keywords solubility prediction machine learning artificial neural network random decision forests

Corresponding Author(s): Jingcai Cheng,Junbo Gong

Online First Date: 12 October 2021 Issue Date: 21 March 2022

Cite this article:

Yiming Ma,Zhenguo Gao,Peng Shi, et al. Machine learning-based solubility prediction and methodology evaluation of active pharmaceutical ingredients in industrial crystallization[J]. Front. Chem. Sci. Eng., 2022, 16(4): 523-535.

URL:

https://academic.hep.com.cn/fcse/EN/10.1007/s11705-021-2083-5
https://academic.hep.com.cn/fcse/EN/Y2022/V16/I4/523

Fig.1 The workflow of the model structure used in the current study.

Tab.1 Statistical results of the RDF and ANN models using molar solubility and log S for the testing set

API names	MW^a) /(g·mol^–1)	Melting point/K	Molar solubility ×10³/(mol·mol^–1)			Log S
API names	MW^a) /(g·mol^–1)	Melting point/K	From refs.^b)	RDF	ANN	From refs.^b)	RDF	ANN
Tridecanedioic acid	244.33	386.210	4.410	7.732	3.769	–2.356	–2.547	–2.453
Pyrimethanil	199.25	369.290	26.700	25.865	35.174	–1.573	–1.621	–1.505
3,4-Dichloronitrobenzene	192.00	316.150	99.670	94.808	86.353	–1.001	–0.906	–1.279
Aceclofenac	354.18	423.150	10.603	13.501	20.326	–1.975	–1.798	–1.957
2,3,4,5-Tetrabromothiophene	399.72	391.150	2.207	4.271	3.418	–2.656	–2.669	–2.626
4,4′-Diaminodiphenylmethane	198.26	363.300	46.100	58.310	32.849	–1.336	–1.336	–1.304
Acetaminophen	151.16	442.350	53.900	50.219	54.240	–1.268	–1.398	–1.540
Famotidine	337.45	439.350	1.700	5.312	0.958	–2.770	–2.719	–3.314
Vanillic acid	168.15	484.900	34.500	34.250	47.182	–1.462	–1.566	–1.419
p-Coumaric acid	164.16	494.350	45.600	50.121	48.328	–1.341	–1.435	–1.551
Flufenamic acid	281.23	407.770	68.100	70.559	72.749	–1.167	–1.223	–1.124
o-Iodoaniline	219.02	325.150	141.338	239.176	198.730	–0.850	–0.724	–1.019
Azoxystrobin	403.39	387.670	1.184	3.487	4.179	–2.927	–2.783	–2.880
Lidocaine	234.33	342.510	470.000	442.998	530.020	–0.328	–0.706	–0.678
1-Naphthoic acid	172.18	436.930	31.090	34.208	32.077	–1.507	–1.566	–1.485
1-(4-Methyl-1-naphthyl)ethanone	216.19	312.000	2.950	4.694	3.543	–2.530	–2.575	–3.033
Tris(3-hydroxypropyl)phosphine oxide	224.24	387.760	22.620	31.759	27.277	–1.646	–1.632	–1.641
Xylitol	152.15	365.150	2.500	7.185	4.186	–2.602	–2.624	–2.624
Antioxidant 626	604.69	448.150	15.011	17.159	18.349	–1.824	–1.646	–1.908
N-Acetyl-L-glutamine	188.18	472.200	0.580	1.240	0.979	–3.237	–3.511	–3.174
Itraconazole	705.64	442.150	0.016	0.114	0.101	–4.801	–3.787	–5.870
Tetrabromo bisphenol A	543.87	454.020	51.070	59.748	40.734	–1.292	–1.404	–1.245
Griseofulvin	352.77	491.610	0.433	1.787	1.391	–3.364	–3.760	–3.304
Trifloxystrobin	408.37	345.670	7.900	11.748	8.638	–2.102	–2.087	–2.021

Tab.2 Prediction model performances in predicting the molar solubility and log S for 24 APIs in the testing set

Tab.3 Summary of statistics for the prediction quality of ANN-log S-output, three LRM models and ANN-QSPR model when applied to the testing set

Fig.2 Summary of the statistical results for the models with MAE in the gray column and SEP in the green column for the testing set.

Tab.4 Performance of the ANN-log S-output model, MSE 01-03, and the ANN-QSPR model for 24 APIs in the testing set

Fig.3 Heatmap between binary variables under molar solubility.

Fig.4 The coefficients of multiple linear regression and stepwise regression.

Fig.5 The vibration in MAE varying with the epochs.

Tab.5 ANN model performances in predicting the molar solubility for different neural network structures

1	H Ma, Y Qu, Z Zhou, S Wang, L Li. Solubility of thiotriazinone in binary solvent mixtures of water+ methanol and water+ ethanol from (283 to 330) K. Journal of Chemical & Engineering Data, 2012, 57(8): 2121–2127 https://doi.org/10.1021/je201149u
2	A Maher, Å Rasmuson, D Croker, B Hodnett. Solubility of the metastable polymorph of piracetam (Form II) in a range of solvents. Journal of Chemical & Engineering Data, 2012, 57(12): 3525–3531 https://doi.org/10.1021/je300711r
3	Y Ma, S Wu, E Macaringue, T Zhang, J Gong, J Wang. Recent progress in continuous crystallization of pharmaceutical products: precise preparation and control. Organic Process Research & Development, 2020, 24(10): 1785–1801 https://doi.org/10.1021/acs.oprd.9b00362
4	Y Wang, S Du, S Wu, L Li, D Zhang, B Yu, L Zhou, H Bekele, J Gong. Thermodynamic and molecular investigation into the solubility, stability and self-assembly of gabapentin anhydrate and hydrate. Journal of Chemical Thermodynamics, 2017, 113: 132–143 https://doi.org/10.1016/j.jct.2017.05.041
5	X Wang, D Zhang, S Liu, Y Chen, L Jia, S Wu. Thermodynamic study of solubility for imatinib mesylate in nine monosolvents and two binary solvent mixtures from 278.15 to 318.15 K. Journal of Chemical & Engineering Data, 2018, 63(11): 4114–4127 https://doi.org/10.1021/acs.jced.8b00551
6	D Kiwala, M Olbrycht, M Balawejder, W Piątkowski, A Seidel-Morgenstern, D Antos. Separation of stereoisomeric mixtures of nafronyl as a representative of compounds possessing two stereogenic centers by coupling crystallization, diastereoisomeric conversion and chromatography. Organic Process Research & Development, 2016, 20(3): 615–625 https://doi.org/10.1021/acs.oprd.5b00361
7	R Qi, J Wang, J Ye, H Hao, Y Bao. The solubility of cefquinome sulfate in pure and mixed solvents. Frontiers of Chemical Science and Engineering, 2016, 10(2): 245–254 https://doi.org/10.1007/s11705-016-1569-z
8	D Herrmannsdörfer, J Stierstorfer, T Klapötke. Solubility behaviour of CL-20 and HMX in organic solvents and solvates of CL-20. Energetic Materials Frontiers, 2021, 2(1): 51–61 https://doi.org/10.1016/j.enmf.2021.01.004
9	S Boobier, D Hose, A Blacker, B Nguyen. Machine learning with physicochemical relationships: solubility prediction in organic solvents and water. Nature Communications, 2020, 11(1): 5753 https://doi.org/10.1038/s41467-020-19594-z
10	Q Cui, S Lu, B Ni, X Zeng, Y Tan, Y Chen, H Zhao. Improved prediction of aqueous solubility of novel compounds by going deeper with deep learning. Frontiers in Oncology, 2020, 10: 121 https://doi.org/10.3389/fonc.2020.00121
11	A Perryman, D Inoyama, J Patel, S Ekins, J Freundlich. Pruned machine learning models to predict aqueous solubility. ACS Omega, 2020, 5(27): 16562–16567 https://doi.org/10.1021/acsomega.0c01251
12	ChemAxon. ChemAxon Website, 2020
13	Y Ran, S Yalkowsky. Prediction of drug solubility by the general solubility equation (GSE). Journal of Chemical Information and Modeling, 2001, 32(22): 354–357
14	D Ellegaard, J Abildskov, J O’Connell. Molecular thermodynamic modeling of mixed solvent solubility. Industrial & Engineering Chemistry Research, 2010, 49(22): 11620–11632 https://doi.org/10.1021/ie101059y
15	W Acree Jr, M Che, G Lee, M Abraham. Calculation of the Abraham model solute descriptors for the pharmaceutical compound acipimox based on experimental solubility data. Physics and Chemistry of Liquids, 2018, 57(3): 382–387 https://doi.org/10.1080/00319104.2018.1467908
16	H Sun, P Shah, K Nguyen, K Yu, E Kerns, M Kabir, Y Wang, X Xu. Predictive models of aqueous solubility of organic compounds built on A large dataset of high integrity. Bioorganic & Medicinal Chemistry, 2019, 27(14): 3110–3114 https://doi.org/10.1016/j.bmc.2019.05.037
17	M Salahinejad, T Le, D Winkler. Aqueous solubility prediction: do crystal lattice interactions help? Molecular Pharmaceutics, 2013, 10(7): 2757–2766 https://doi.org/10.1021/mp4001958
18	S Chinta, R Rengaswamy. Machine learning derived quantitative structure property relationship (QSPR) to predict drug solubility in binary solvent systems. Industrial & Engineering Chemistry Research, 2019, 58(8): 3082–3092 https://doi.org/10.1021/acs.iecr.8b04584
19	S Fioressi, D Bacelo, C Rojas, J Aranda, P Duchowicz. Conformation-independent quantitative structure-property relationships study on water solubility of pesticides. Ecotoxicology and Environmental Safety, 2019, 171: 47–53 https://doi.org/10.1016/j.ecoenv.2018.12.056
20	O Wahab, L Olasunkanmi, K Govender, P Govender. Prediction of aqueous solubility by treatment of COSMO-RS data with empirical solubility equations: the roles of global orbital cut-off and COSMO solvent radius. Theoretical Chemistry Accounts, 2019, 138(6): 80 https://doi.org/10.1007/s00214-019-2470-x
21	D Abranches, J Benfica, S Shimizu, J Coutinho. Solubility enhancement of hydrophobic substances in water/cyrene mixtures: a computational study. Industrial & Engineering Chemistry Research, 2020, 59(40): 18247–18253 https://doi.org/10.1021/acs.iecr.0c03155
22	E Modarresi, J Abildskov, R Gani, P Crafts. Model-based calculation of solid solubility for solvent selections: a review. Industrial & Engineering Chemistry Research, 2008, 47(15): 5234–5242 https://doi.org/10.1021/ie0716363
23	C Shang, F You. Data analytics and machine learning for smart process manufacturing: recent advances and perspectives in the big data era. Engineering, 2019, 5(6): 1010–1016 https://doi.org/10.1016/j.eng.2019.01.019
24	Y Xie, C Zhang, X Hu, C Zhang, S Kelley, J Atwood, J Lin. Machine learning assisted synthesis of metal-organic nanocapsules. Journal of the American Chemical Society, 2020, 142(3): 1475–1481 https://doi.org/10.1021/jacs.9b11569
25	Y Dong, C Wu, C Zhang, Y Liu, J Cheng, J Lin. Bandgap prediction by deep learning in configurationally hybridized graphene and boron nitride. npj Computational Materials, 2019, 5, 26
26	D Xin, N Gonnella, X He, K Horspool. Solvate prediction for pharmaceutical organic molecules with machine learning. Crystal Growth & Design, 2019, 19(3): 1903–1911 https://doi.org/10.1021/acs.cgd.8b01883
27	A Ghosh, L Louis, K Arora, B Hancock, J Krzyzaniak, P Meenan, S Nakhmanson, G Wood. Assessment of machine learning approaches for predicting the crystallization propensity of active pharmaceutical ingredients. CrystEngComm, 2019, 21(8): 1215–1223 https://doi.org/10.1039/C8CE01589A
28	W Paengjuntuek, L Thanasinthana, A Arpornwichanop. Neural network-based optimal control of a batch crystallizer. Neurocomputing, 2012, 83: 158–164 https://doi.org/10.1016/j.neucom.2011.12.008
29	D Han, T Karmakar, Z Bjelobrk, J Gong, M Parrinello. Solvent-mediated morphology selection of the active pharmaceutical ingredient isoniazid: experimental and simulation studies. Chemical Engineering Science, 2018, 204: 320–328 https://doi.org/10.1016/j.ces.2018.10.022
30	N Wang, X Huang, H Gong, Y Zhou, X Li, F Li, Y Bao, C Xie, Z Wang, Q Yin, H Hao. Thermodynamic mechanism of selective cocrystallization explored by MD simulation and phase diagram analysis. AIChE Journal. American Institute of Chemical Engineers, 2019, 65(5): e16570 https://doi.org/10.1002/aic.16570
31	Y Ma, Y Cao, Y Yang, W Li, P Shi, S Wang, W Tang. Thermodynamic analysis and molecular dynamic simulation of the solubility of vortioxetine hydrobromide in three binary solvent mixtures. Journal of Molecular Liquids, 2018, 272: 676–688 https://doi.org/10.1016/j.molliq.2018.09.130
32	T Zhang, Z Li, Y Wang, C Li, B Yu, X Zheng, L Jiang, J Gong. Determination and correlation of solubility and thermodynamic properties of l-methionine in binary solvents of water+ (methanol, ethanol, acetone). Journal of Chemical Thermodynamics, 2016, 96: 82–92 https://doi.org/10.1016/j.jct.2015.12.022
33	A Raudino, M Sarpietro, M Pannuzzo. Differential scanning calorimetry (DSC): theoretical fundamentals. In: Drug-Biomembrane Interaction Studies. Pignatello R, ed. Cambridge, UK: Woodhead Publishing Limited, 2013: 127–168
34	G Foca, A Marchetti, L Tassi, A Ulrici. Modelling of experimental thermophysical data by mixing of a ternary solvent system. Solution Chemistry Research Progress, 2011: 5–49
35	S Price, J. Brandenburg Molecular crystal structure prediction. Non-Covalent Interactions in Quantum Chemistry and Physics, 2017, 333–363
36	P Shi, Y Ma, D Han, S Du, T Zhang, Z Li. Uncovering the solubility behavior of vitamin B6 hydrochloride in three aqueous binary solvents by thermodynamic analysis and molecular dynamic simulation. Journal of Molecular Liquids, 2019, 283: 584–595 https://doi.org/10.1016/j.molliq.2019.03.082
37	S Zhao, Y Ma, W Tang. Thermodynamic analysis and molecular dynamic simulation of solid-liquid phase equilibrium of griseofulvin in three binary solvent systems. Journal of Molecular Liquids, 2019, 294: 111600 https://doi.org/10.1016/j.molliq.2019.111600
38	G Rong, A Mendez, E Bou Assi, B Zhao, M Sawan. Artificial intelligence in healthcare: review and prediction case studies. Engineering, 2020, 6(3): 291–301 https://doi.org/10.1016/j.eng.2019.08.015
39	J Vegh. How Amdahl’s Law limits the performance of large artificial neural networks: why the functionality of full-scale brain simulation on processor-based simulators is limited. Brain Informatics, 2019, 6(1): 4 https://doi.org/10.1186/s40708-019-0097-2
40	J Xu, Y Chen, T Xie, X Zhao, B Xiong, Z Chen. Prediction of triaxial behavior of recycled aggregate concrete using multivariable regression and artificial neural network techniques. Construction & Building Materials, 2019, 226: 534–554 https://doi.org/10.1016/j.conbuildmat.2019.07.155
41	F Rosenblatt. The perception: a probabilistic model for information storage and organization in the brain. Psychological Review, 1988, 65(6): 89–114
42	J McDonagh, N Nath, L De Ferrari, T van Mourik, J Mitchell. Uniting cheminformatics and chemical theory to predict the intrinsic aqueous solubility of crystalline druglike molecules. Journal of Chemical Information and Modeling, 2014, 54(3): 844–856 https://doi.org/10.1021/ci4005805
43	B Rizkin, R Hartman. Supervised machine learning for prediction of zirconocene-catalyzed α-olefin polymerization. Chemical Engineering Science, 2019, 210: 115224 https://doi.org/10.1016/j.ces.2019.115224
44	L Breiman. Random forests. Machine Learning, 2001, 45(1): 5–32 https://doi.org/10.1023/A:1010933404324
45	T Ho. decision forest Random. In: Proceedings of 3rd International Conference on Document Analysis and Recongnition. Montreal, Canada, 1995, 278–282
46	S Lee, J Kim, N Moon. Random forest and WiFi fingerprint-based indoor location recognition system using smart watch. Human-centric Computing and Information Sciences, 2019, 9(1): 6 https://doi.org/10.1186/s13673-019-0168-7
47	F de Santana, W Borges Neto, R Poppi. Random forest as one-class classifier and infrared spectroscopy for food adulteration detection. Food Chemistry, 2019, 293: 323–332 https://doi.org/10.1016/j.foodchem.2019.04.073
48	T Zhou, X Sun, X Xia, B Li, X Chen. Improving defect prediction with deep forest. Information and Software Technology, 2019, 114: 204–216 https://doi.org/10.1016/j.infsof.2019.07.003
49	A Tarasova, F Burden, J Gasteiger, D Winkler. Robust modelling of solubility in supercritical carbon dioxide using Bayesian methods. Journal of Molecular Graphics & Modelling, 2010, 28(7): 593–597 https://doi.org/10.1016/j.jmgm.2009.12.004
50	T Le, V Epa, F Burden, D Winkler. Quantitative structure-property relationship modeling of diverse materials properties. Chemical Reviews, 2012, 112(5): 2889–2919 https://doi.org/10.1021/cr200066h
51	A Clark, P Labute. Detection and assignment of common scaffolds in project databases of lead molecules. Journal of Medicinal Chemistry, 2009, 52(2): 469–483 https://doi.org/10.1021/jm801098a
52	Molecular Operating Environment (MOE). Version 2019.0102. Montreal: Chemical Computing Group ULC, 2019

[1]

FCE-21017-OF-MY_suppl_1

Download

[1]	Quirin Göttl, Dominik G. Grimm, Jakob Burger. Automated synthesis of steady-state continuous processes using reinforcement learning[J]. Front. Chem. Sci. Eng., 2022, 16(2): 288-302.
[2]	Haoqin Fang, Jianzhao Zhou, Zhenyu Wang, Ziqi Qiu, Yihua Sun, Yue Lin, Ke Chen, Xiantai Zhou, Ming Pan. Hybrid method integrating machine learning and particle swarm optimization for smart chemical process operations[J]. Front. Chem. Sci. Eng., 2022, 16(2): 274-287.
[3]	Ewan Chee, Wee Chin Wong, Xiaonan Wang. An integrated approach for machine-learning-based system identification of dynamical systems under control: application towards the model predictive control of a highly nonlinear reactor system[J]. Front. Chem. Sci. Eng., 2022, 16(2): 237-250.
[4]	Patrick Otto Ludl, Raoul Heese, Johannes Höller, Norbert Asprion, Michael Bortz. Using machine learning models to explore the solution space of large nonlinear systems underlying flowsheet simulations with constraints[J]. Front. Chem. Sci. Eng., 2022, 16(2): 183-197.
[5]	PAN Yong, JIANG Juncheng, WANG Zhirong. Prediction of the flash points of alkanes by group bond contribution method using artificial neural networks[J]. Front. Chem. Sci. Eng., 2007, 1(4): 390-394.

Viewed

Full text

Abstract

Cited

Shared

Discussed