贝叶斯优化方法在需钠弧菌生产1，3-丙二醇中的应用

doi:10.11949/0438-1157.20221102

摘要/Abstract

摘要：

1，3-丙二醇是一种重要的化工原料，生物法发酵制备1，3-丙二醇具有操作简便、反应条件温和、副产物少等优点，使用需钠弧菌作为新的工业底盘细胞生产1，3-丙二醇有很好的应用前景。高产菌株的构建过程中优化基因强度组合时，具有单次实验周期长、成本高且体系复杂、非线性强的特点。为了实现对1，3-丙二醇合成途径关键基因强度组合的快速优化，使用了一种以高斯过程回归算法为代理模型，以增益期望为采集函数的贝叶斯优化方法。在每轮迭代中，高斯过程回归算法用于拟合当前数据并预测未知点的概率分布，增益期望函数将概率分布映射到解空间中，选择解空间中最大值对应的点作为下一轮的实验点，实验后进入下一轮迭代。构建高产1，3-丙二醇的需钠弧菌过程中使用贝叶斯优化方法优化关键基因强度组合，在三轮迭代后搜索到了最优的基因强度组合，1，3-丙二醇的产量达到（13.01±0.63）g/L，较第一组实验点中最高值提高了8.32%。

关键词: 1, 3-丙二醇, 需钠弧菌, 贝叶斯优化, 优化设计, 生物过程

Abstract:

1,3-Propanediol is an important chemical raw material. The preparation of 1,3-propanediol by biological fermentation has the advantages of simple operation, mild reaction conditions and few by-products. The use of Vibrio natriegens as a new industrial chassis to produce 1,3-propanediol has important application prospects, but the construction of high-yielding strains has the characteristics of long single experiment period, high cost, and strong nonlinearity. In order to fast optimize the combination of key gene intensity of 1,3-propanediol synthesis pathway, we use a Bayesian optimization method with Gaussian process regression algorithm as surrogate model and expected improvement as acquisition function. In each iteration, the Gaussian process regression algorithm is used to fit the current data and predict the probability distribution of unknown points. The expected improvement function maps the probability distribution to the solution space, selects the point corresponding to the maximum value in the solution space as the experimental point in the next iteration, and enters the next iteration after experiment. The Bayesian optimization method was used to optimize the key gene intensity combination in the process of constructing a high-yielding 1,3-propanediol-producing Vibrio natriegens. After three iterations, the optimal gene strength combination was searched. The 1,3-propanediol yield reached (13.01±0.63) g/L, which was 8.32% higher than the highest value in the first group of test points.

Key words: 1,3-propanediol, Vibrio natriegens, Bayesian optimization, optimal design, bioprocess

中图分类号:

TQ 033

黄新烨, 张冶, 张书源, 陈振, 邱彤. 贝叶斯优化方法在需钠弧菌生产1，3-丙二醇中的应用[J]. 化工学报, 2022, 73(11): 5039-5046.

Xinye HUANG, Ye ZHANG, Shuyuan ZHANG, Zhen CHEN, Tong QIU. Application of Bayesian optimization method in the production of 1,3-propanediol by Vibrio natriegens[J]. CIESC Journal, 2022, 73(11): 5039-5046.

图/表 9

图 1 需钠弧菌合成1,3-丙二醇途径

Fig.1 Synthesis of 1,3-PDO by Vibrio natriegens

图2 基因启动子调控示意图

Fig.2 Schematic diagram of gene promoter regulation

表 1 本研究所采用的组成型启动子［16］

Table 1 Constitutive promoters used in this study［16］

No.	核苷酸序列（5'→3'）	相对强度
1	CTTATGGCTAGCTCAGTCCTAGGGACAGTGCTAGC	0.024
2	TTTACGGCTAGCTCAGTCCTAGGGATAGTGCTAGC	0.045
3	CTGACGGCTAGCTCAGTCCTAGGGATAGTGCTAGC	0.065
4	TTGATGGCTAGCTCAGTCCTAGGGATTATGCTAGC	0.078
5	TTGATGGCTAGCTCAGTCCTAGGTACAGTGCTAGC	0.116
6	TTGATGGCTAGCTCAGTCCTAGGTATTGTGCTAGC	0.131
7	TTGATGGCTAGCTCAGTCCTAGGTACTATGCTAGC	0.147
8	TTGACGGCTAGCTCAGTCCTAGGTACTGTGCTAGC	0.192
9	TTGATGGCTAGCTCAGTCCTAGGTACAATGCTAGC	0.235
10	TTGATGGCTAGCTCAGTCCTAGGTATAGTGCTAGC	0.265
11	TTGACGGCTAGCTCAGTCCTAGGTATTGTGCTAGC	0.297
12	TTGATGGCTAGCTCAGTCCTAGGTATAATGCTAGC	0.371
13	TTGACGGCTAGCTCAGTCCTAGGTACAGTGCTAGC	0.457
14	TTGACAGCTAGCTCAGTCCTAGGTATTGTGCTAGC	0.645
15	TTGACAGCTAGCTCAGTCCTAGGTATAATGCTAGC	1.000

图3 用于实验条件优选的贝叶斯优化算法框图

Fig.3 Bayesian optimization algorithm for Vibrio natriegens

表2 需钠弧菌合成1，3-丙二醇的初始实验数据

Table 2 Raw data of 1，3-propanediol synthesis by Vibrio natriegens

$d h a B C E$	$y q h D$	$l n (d h a B C E)$	$l n (y q h D)$	12 h 1,3-PDO/(g/L)
0.024	0.116	-3.730	-2.154	1.13±0.80
0.147	0.065	-1.917	-2.733	0.41±0.99
0.192	0.045	-1.650	-3.101	3.20±0.75
0.045	0.192	-3.101	-1.650	0.15±1.16
0.131	0.265	-2.033	-1.328	2.97±0.13
0.116	0.131	-2.154	-2.033	3.78±0.39
0.235	0.457	-1.448	-0.783	8.22±0.69
0.235	0.371	-1.448	-0.992	7.54±0.83
0.265	0.192	-1.328	-1.650	6.31±0.55
0.371	0.192	-0.992	-1.650	3.15±0.16
0.371	0.645	-0.992	-0.439	7.31±0.99
0.645	0.045	-0.439	-3.101	0.90±0.03
0.024	1.000	-3.730	0	0.08±0.01
0.371	0.457	-0.992	-0.783	3.05±0.56
0.291	1.000	-1.234	0	12.01±0.33

表2 需钠弧菌合成1，3-丙二醇的初始实验数据

Table 2 Raw data of 1，3-propanediol synthesis by Vibrio natriegens

$d h a B C E$	$y q h D$	$l n (d h a B C E)$	$l n (y q h D)$	12 h 1,3-PDO/(g/L)
0.024	0.116	-3.730	-2.154	1.13±0.80
0.147	0.065	-1.917	-2.733	0.41±0.99
0.192	0.045	-1.650	-3.101	3.20±0.75
0.045	0.192	-3.101	-1.650	0.15±1.16
0.131	0.265	-2.033	-1.328	2.97±0.13
0.116	0.131	-2.154	-2.033	3.78±0.39
0.235	0.457	-1.448	-0.783	8.22±0.69
0.235	0.371	-1.448	-0.992	7.54±0.83
0.265	0.192	-1.328	-1.650	6.31±0.55
0.371	0.192	-0.992	-1.650	3.15±0.16
0.371	0.645	-0.992	-0.439	7.31±0.99
0.645	0.045	-0.439	-3.101	0.90±0.03
0.024	1.000	-3.730	0	0.08±0.01
0.371	0.457	-0.992	-0.783	3.05±0.56
0.291	1.000	-1.234	0	12.01±0.33

图4 高斯过程回归模型参数调整

Fig.4 Parameter adjustment of Gaussian process regression

图5 贝叶斯优化详细迭代过程示意图1、2、3分别表示第一次迭代、第二次迭代和第三次迭代的结果;背景颜色表示预测均值;红色的点为第一次迭代的推荐方案,青色的点为第二次的推荐方案,橘色的点为第三次的推荐方案,分别记为预测点1、预测点2、预测点3

Fig.5 Detailed iterative process of Bayesian optimization

图6 需钠弧菌生产1,3-丙二醇优化结果

Fig.6 Optimization of 1,3-propanediol production by Vibrio natriegens

表 3 迭代过程预测点结果

Table 3 Prediction point results of iterative process

预测点	预测均值	预测方差	实际值	训练集RMSE
1	12.38	0.262	12.49	1.84×10^-9
2	11.49	0.248	13.01	1.66×10^-9
3	14.56	0.230	9.51	2.83×10^-9

参考文献 20

1	朱丙田, 刘德华, 任海玉, 等. 1,3-丙二醇发酵条件的探索[J]. 化工冶金, 2000(4): 420-422.
	Zhu B T, Liu D H, Ren H Y, et al. Experiments on optimal conditions for 1,‍3-propanediol fermentation[J]. Engineering Chemistry ＆ Metallurgy, 2000(4): 420-422.
2	李晓姝, 张霖, 高大成, 等. 发酵法生产1,3-丙二醇的研究进展[J]. 化工进展, 2017, 36(4): 1395-1403.
	Li X S, Zhang L, Gao D C, et al. Progress on the production of 1, 3-propanediol by fermentation[J]. Chemical Industry and Engineering Progress, 2017, 36(4): 1395-1403.
3	谢家明, 徐泽辉, 夏蓉晖, 等. 1,3-丙二醇制备工艺的研究进展[J]. 合成纤维, 2005, 34(2): 13-16, 48.
	Xie J M, Xu Z H, Xia R H, et al. Research progress on the technology for producing 1,3-propanediol[J]. Synthetic Fiber in China, 2005, 34(2): 13-16, 48.
4	刘德华, 刘宏娟, 程可可. 微生物发酵法生产1,3-丙二醇研究进展[J]. 合成纤维, 2005, 34(9): 11-15.
	Liu D H, Liu H J, Cheng K K. Research progress on the production of 1,3-propanediol by fermentation[J]. Synthetic Fiber in China, 2005, 34(9): 11-15.
5	Huang H, Gong C S, Tsao G T. Production of 1,3-propanediol by Klebsiella pneumoniae [J]. Applied Biochemistry and Biotechnology, 2002, 98/99/100: 687-698.
6	陈振, 刘宏娟, 刘德华. 有氧条件下Klebsiella pneumoniae发酵生产1, 3-丙二醇的研究[J]. 现代化工, 2006, 26(S2): 297-300.
	Chen Z, Liu H J, Liu D H. Study on 1,3-propanediol production by Klebsiella pneumoniae under aerobic conditions[J]. Modern Chemical Industry, 2006, 26(S2): 297-300.
7	Zhang Y, Li Z H, Liu Y, et al. Systems metabolic engineering of Vibrio natriegens for the production of 1,3-propanediol[J]. Metabolic Engineering, 2021, 65: 52-65.
8	Jalasutram V, Jetty A. Optimization of 1,3-propanediol production by Klebsiella pneumoniae 141B using Taguchi methodology: improvement in production by cofermentation studies[J]. Research in Biotechnology, 2011, 2(2): 189-197.
9	Xu Q, Yang X T, Liu C W, et al. Chemical plant flare minimization via plantwide dynamic simulation[J]. Industrial & Engineering Chemistry Research, 2009, 48(7): 3505-3512.
10	Zhang A H, Zhu K Y, Zhuang X Y, et al. A robust soft sensor to monitor 1, 3-propanediol fermentation process by Clostridium butyricum based on artificial neural network[J]. Biotechnology and Bioengineering, 2020, 117(11): 3345-3355.
11	Zhou Y K, Li G, Dong J K, et al. MiYA, an efficient machine-learning workflow in conjunction with the YeastFab assembly strategy for combinatorial optimization of heterologous metabolic pathways in Saccharomyces cerevisiae [J]. Metabolic Engineering, 2018, 47: 294-302.
12	李浩然, 邱彤. 基于因果分析的烧结生产状态预测模型[J]. 化工学报, 2021, 72(3): 1438-1446.
	Li H R, Qiu T. Sintering production state prediction model based on causal analysis[J]. CIESC Journal, 2021, 72(3): 1438-1446.
13	Mockus J. Bayesian Approach to Global Optimization: Theory and Applications[M/OL]. Netherlands: Springer, 1989[2021-07-12]. .
14	Griffiths R R, Hernández-Lobato J M. Constrained Bayesian optimization for automatic chemical design using variational autoencoders[J]. Chemical Science, 2019, 11(2): 577-586.
15	Gómez-Bombarelli R, Wei J N, Duvenaud D, et al. Automatic chemical design using a data-driven continuous representation of molecules[J]. ACS Central Science, 2018, 4(2): 268-276.
16	Iverson S V, Haddock T L, Beal J, et al. CIDAR MoClo: improved MoClo assembly standard and new E. coli part library enable rapid combinatorial design for synthetic and traditional biology[J]. ACS Synthetic Biology, 2016, 5(1): 99-103.
17	Shahriari B, Swersky K, Wang Z Y, et al. Taking the human out of the loop: a review of Bayesian optimization[J]. Proceedings of the IEEE, 2016, 104(1): 148-175.
18	Brochu E, Cora V M, de Freitas N. A tutorial on Bayesian optimization of expensive cost functions, with application to active user modeling and hierarchical reinforcement learning[EB/OL]. 2010. .
19	Rasmussen C E, Williams C K I. Gaussian Processes for Machine Learning[M]. Cambridge: The MIT Press, 2006.
20	Bull A D. Convergence rates of efficient global optimization algorithms[J]. Journal of Machine Learning Research, 2011, 12(10): 2879-2904.

[1]	吴曦, 区祖迪, 张鑫杰, 徐士鸣, 朱晓静. HFO-1243zf爆燃特性实验研究[J]. 化工学报, 2023, 74(S1): 346-352.
[2]	陈爱强, 代艳奇, 刘悦, 刘斌, 吴翰铭. 基板温度对HFE7100液滴蒸发过程的影响研究[J]. 化工学报, 2023, 74(S1): 191-197.
[3]	陈哲文, 魏俊杰, 张玉明. 超临界水煤气化耦合SOFC发电系统集成及其能量转化机制[J]. 化工学报, 2023, 74(9): 3888-3902.
[4]	齐聪, 丁子, 余杰, 汤茂清, 梁林. 基于选择吸收纳米薄膜的太阳能温差发电特性研究[J]. 化工学报, 2023, 74(9): 3921-3930.
[5]	文兆伦, 李沛睿, 张忠林, 杜晓, 侯起旺, 刘叶刚, 郝晓刚, 官国清. 基于自热再生的隔壁塔深冷空分工艺设计及优化[J]. 化工学报, 2023, 74(7): 2988-2998.
[6]	江锦波, 彭新, 许文烜, 门日秀, 刘畅, 彭旭东. 泵出型螺旋槽油气密封泄漏特性及参数影响研究[J]. 化工学报, 2023, 74(6): 2538-2554.
[7]	徐文超, 孙志高, 李翠敏, 李娟, 黄海峰. 静态条件下表面活性剂E-1310对HCFC-141b水合物生成的影响[J]. 化工学报, 2023, 74(5): 2179-2185.
[8]	孙永尧, 高秋英, 曾文广, 王佳铭, 陈艺飞, 周永哲, 贺高红, 阮雪华. 面向含氮油田伴生气提质利用的膜耦合分离工艺设计优化[J]. 化工学报, 2023, 74(5): 2034-2045.
[9]	刘尚豪, 贾胜坤, 罗祎青, 袁希钢. 基于梯度提升决策树的三组元精馏流程结构最优化[J]. 化工学报, 2023, 74(5): 2075-2087.
[10]	周必茂, 许世森, 王肖肖, 刘刚, 李小宇, 任永强, 谭厚章. 烧嘴偏转角度对气化炉渣层分布特性的影响[J]. 化工学报, 2023, 74(5): 1939-1949.
[11]	王泽栋, 石至平, 刘丽艳. 考虑气泡非均匀耗散的矩形反应器声流场数值模拟及结构优化[J]. 化工学报, 2023, 74(5): 1965-1973.
[12]	许文烜, 江锦波, 彭新, 门日秀, 刘畅, 彭旭东. 宽速域三种典型型槽油气密封泄漏与成膜特性对比研究[J]. 化工学报, 2023, 74(4): 1660-1679.
[13]	李纪元, 李金旺, 周刘伟. 不同扰流结构冷板传热性能研究[J]. 化工学报, 2023, 74(4): 1474-1488.
[14]	陈俊先, 姬忠礼, 赵瑜, 张倩, 周岩, 刘猛, 刘震. 基于微波技术的天然气管道内颗粒物在线检测方法研究[J]. 化工学报, 2023, 74(3): 1042-1053.
[15]	王锋, 陈钰, 裴鸿艳, 刘东东, 张静, 张立新. 1，2，4-𫫇二唑类衍生物的设计、合成及抗菌活性[J]. 化工学报, 2023, 74(3): 1390-1398.