Coptis chinensis(Huanglian) is a commonly used traditional Chinese medicine(TCM) herb and alkaloids are the most important chemical constituents in it. In the present study, an isocratic reverse phase high performance...Coptis chinensis(Huanglian) is a commonly used traditional Chinese medicine(TCM) herb and alkaloids are the most important chemical constituents in it. In the present study, an isocratic reverse phase high performance liquid chromatography(RP-HPLC) method allowing the separation of six alkaloids in Huanglian was for the first time developed under the quality by design(Qb D) principles. First, five chromatographic parameters were identified to construct a Plackett-Burman experimental design. The critical resolution, analysis time, and peak width were responses modeled by multivariate linear regression. The results showed that the percentage of acetonitrile, concentration of sodium dodecyl sulfate, and concentration of potassium phosphate monobasic were statistically significant parameters(P < 0.05). Then, the Box-Behnken experimental design was applied to further evaluate the interactions between the three parameters on selected responses. Full quadratic models were built and used to establish the analytical design space. Moreover, the reliability of design space was estimated by the Bayesian posterior predictive distribution. The optimal separation was predicted at 40% acetonitrile, 1.7 g·m L-1of sodium dodecyl sulfate and 0.03 mol·m L-1 of potassium phosphate monobasic. Finally, the accuracy profile methodology was used to validate the established HPLC method. The results demonstrated that the Qb D concept could be efficiently used to develop a robust RP-HPLC analytical method for Huanglian.展开更多
This study presents a Bayesian methodology for de- signing step stress accelerated degradation testing （SSADT） and its application to batteries. First, the simulation-based Bayesian de- sign framework for SSADT is p...This study presents a Bayesian methodology for de- signing step stress accelerated degradation testing （SSADT） and its application to batteries. First, the simulation-based Bayesian de- sign framework for SSADT is presented. Then, by considering his- torical data, specific optimal objectives oriented Kullback-Leibler （KL） divergence is established. A numerical example is discussed to illustrate the design approach. It is assumed that the degrada- tion model （or process） follows a drift Brownian motion; the accele- ration model follows Arrhenius equation; and the corresponding parameters follow normal and Gamma prior distributions. Using the Markov Chain Monte Carlo （MCMC） method and WinBUGS software, the comparison shows that KL divergence is better than quadratic loss for optimal criteria. Further, the effect of simulation outiiers on the optimization plan is analyzed and the preferred sur- face fitting algorithm is chosen. At the end of the paper, a NASA lithium-ion battery dataset is used as historical information and the KL divergence oriented Bayesian design is compared with maxi- mum likelihood theory oriented locally optimal design. The results show that the proposed method can provide a much better testing plan for this engineering application.展开更多
This paper discussed Bayesian variable selection methods for models from split-plot mixture designs using samples from Metropolis-Hastings within the Gibbs sampling algorithm. Bayesian variable selection is easy to im...This paper discussed Bayesian variable selection methods for models from split-plot mixture designs using samples from Metropolis-Hastings within the Gibbs sampling algorithm. Bayesian variable selection is easy to implement due to the improvement in computing via MCMC sampling. We described the Bayesian methodology by introducing the Bayesian framework, and explaining Markov Chain Monte Carlo (MCMC) sampling. The Metropolis-Hastings within Gibbs sampling was used to draw dependent samples from the full conditional distributions which were explained. In mixture experiments with process variables, the response depends not only on the proportions of the mixture components but also on the effects of the process variables. In many such mixture-process variable experiments, constraints such as time or cost prohibit the selection of treatments completely at random. In these situations, restrictions on the randomisation force the level combinations of one group of factors to be fixed and the combinations of the other group of factors are run. Then a new level of the first-factor group is set and combinations of the other factors are run. We discussed the computational algorithm for the Stochastic Search Variable Selection (SSVS) in linear mixed models. We extended the computational algorithm of SSVS to fit models from split-plot mixture design by introducing the algorithm of the Stochastic Search Variable Selection for Split-plot Design (SSVS-SPD). The motivation of this extension is that we have two different levels of the experimental units, one for the whole plots and the other for subplots in the split-plot mixture design.展开更多
Reliability-based design (RBD) is being adopted by geotechnical design codes worldwide, and it is therefore necessary that rock engineering practice evolves to embrace RBD. This paper examines the Hoek-Brown (H-B) str...Reliability-based design (RBD) is being adopted by geotechnical design codes worldwide, and it is therefore necessary that rock engineering practice evolves to embrace RBD. This paper examines the Hoek-Brown (H-B) strength criterion within the RBD framework, and presents three distinct analyses using a Bayesian approach. Firstly, a compilation of intact compressive strength test data for six rock types is used to examine uncertainty and variability in the estimated H-B parameters m and σc, and corresponding predicted axial strength. The results suggest that within- and between-rock type variabilities are so large that these parameters need to be determined from rock testing campaigns, rather than reference values being used. The second analysis uses an extensive set of compressive and tensile (both direct and indirect) strength data for a granodiorite, together with a new Bayesian regression model, to develop joint probability distributions of m and σc suitable for use in RBD. This analysis also shows how compressive and indirect tensile strength data may be robustly used to fit an H-B criterion. The third analysis uses the granodiorite data to investigate the important matter of developing characteristic strength criteria. Using definitions from Eurocode 7, a formal Bayesian interpretation of characteristic strength is proposed and used to analyse strength data to generate a characteristic criterion. These criteria are presented in terms of characteristic parameters mk and σck, the values of which are shown to depend on the testing regime used to obtain the strength data. The paper confirms that careful use of appropriate Bayesian statistical analysis allows the H-B criterion to be brought within the framework of RBD. It also reveals that testing guidelines such as the International Society for Rock Mechanics and Rock Engineering (ISRM) suggested methods will require modification in order to support RBD. Importantly, the need to fully understand the implications of uncertainty in nonlinear strength criteria is identified.展开更多
Soil-water characteristic curve (SWCC) is significant to estimate the site-specific unsaturated soil properties (such as unsaturated shear strength and coefficient of permeability) for geotechnical analyses involving ...Soil-water characteristic curve (SWCC) is significant to estimate the site-specific unsaturated soil properties (such as unsaturated shear strength and coefficient of permeability) for geotechnical analyses involving unsaturated soils. Determining SWCC can be achieved by fitting data points obtained according to the prescribed experimental scheme, which is specified by the number of measuring points and their corresponding values of the control variable. The number of measuring points is limited since direct measurement of SWCC is often costly and time-consuming. Based on the limited number of measuring points, the estimated SWCC is unavoidably associated with uncertainties, which depends on measurement data obtained from the prescribed experimental scheme. Therefore, it is essential to plan the experimental scheme so as to reduce the uncertainty in the estimated SWCC. This study presented a Bayesian approach, called OBEDO, for probabilistic experimental design optimization of measuring SWCC based on the prior knowledge and information of testing apparatus. The uncertainty in estimated SWCC is quantified and the optimal experimental scheme with the maximum expected utility is determined by Subset Simulation optimization (SSO) in candidate experimental scheme space. The proposed approach is illustrated using an experimental design example given prior knowledge and the information of testing apparatus and is verified based on a set of real loess SWCC data, which were used to generate random experimental schemes to mimic the arbitrary arrangement of measuring points during SWCC testing in practice. Results show that the arbitrary arrangement of measuring points of SWCC testing is hardly superior to the optimal scheme obtained from OBEDO in terms of the expected utility. The proposed OBEDO approach provides a rational tool to optimize the arrangement of measuring points of SWCC test so as to obtain SWCC measurement data with relatively high expected utility for uncertainty reduction.展开更多
Finding the optimal dose combination in two-agent dose-finding trials is challenging due to limited sample sizes and the extensive range of potential doses.Unlike traditional chemotherapy or radiotherapy,which primari...Finding the optimal dose combination in two-agent dose-finding trials is challenging due to limited sample sizes and the extensive range of potential doses.Unlike traditional chemotherapy or radiotherapy,which primarily focuses on identifying the maximum tolerated dose(MTD),therapies involving targeted and immune agents facilitate the identifica-tion of an optimal biological dose combination(OBDC)by simultaneously evaluating both toxicity and efficacy.Cur-rently,most approaches to determining the OBDC in the literature are model-based and require complex model fittings,making them cumbersome and challenging to implement.To address these challenges,we developed a novel model-as-sisted approach called uTPI-Comb.This approach refines the established utility-based toxicity probability interval design by integrating a strategically devised zone-based local and global candidate set searching strategy,which can effectively optimize the decision-making process for two-agent dose escalation or de-escalation in drug combination trials.Extensive simulation studies demonstrate that the uTPI-Comb design speeds up the dose-searching process and provides substantial improvements over existing model-based methods in determining the optimal biological dose combinations.展开更多
Creating new molecules with desired properties is a fundamental and challenging problem in chemistry. Reinforcement learning (RL) has shown its utility in this area where the target chemical property values can serve ...Creating new molecules with desired properties is a fundamental and challenging problem in chemistry. Reinforcement learning (RL) has shown its utility in this area where the target chemical property values can serve as a reward signal. At each step of making a new molecule, the RL agent learns selecting an action from a list of many chemically valid actions for a given molecule, implying a great uncertainty associated with its learning. In a traditional implementation of deep RL algorithms, deterministic neural networks are typically employed, thus allowing the agent to choose one action from one sampled action at each step. In this paper, we proposed a new strategy of applying Bayesian neural networks to RL to reduce uncertainty so that the agent can choose one action from a pool of sampled actions at each step, and investigated its benefits in molecule design. Our experiments suggested the Bayesian approach could create molecules of desirable chemical quality while maintained their diversity, a very difficult goal to achieve in machine learning of molecules. We further exploited their diversity by using them to train a generative model to yield more novel drug-like molecules, which were absent in the training molecules as we know novelty is essential for drug candidate molecules. In conclusion, Bayesian approach could offer a balance between exploitation and exploration in RL, and a balance between optimization and diversity in molecule design.展开更多
By analyzing the shortage of reliability test design and thinking over the producer's risk and consumer's risk, the information fusion technology is used to set up a reliability test design model( RTDM). By an...By analyzing the shortage of reliability test design and thinking over the producer's risk and consumer's risk, the information fusion technology is used to set up a reliability test design model( RTDM). By analyzing the demands and constraint conditions of the RTDM and with applications of Bayesian approach and Monte Carlo method( MCM),this paper puts forward the exponential distributed subsystems and the information fusion technology among them. According to the posteriori risk criteria,formulas of producer's risk and consumer's risk were also inferred,and with the help of Matlab software,selection of the optimum test plan was solved. Finally,validity of the model had been proved by a test of series parallel system.展开更多
The precise and accurate knowledge of genetic parameters is a prerequisite for making efficient selection strategies in breeding programs.A number of estimators of heritability about important economic traits in many ...The precise and accurate knowledge of genetic parameters is a prerequisite for making efficient selection strategies in breeding programs.A number of estimators of heritability about important economic traits in many marine mollusks are available in the literature,however very few research have evaluated about the accuracy of genetic parameters estimated with different family structures.Thus,in the present study,the effect of parent sample size for estimating the precision of genetic parameters of four growth traits in clam M.meretrix by factorial designs were analyzed through restricted maximum likelihood（REML） and Bayesian.The results showed that the average estimated heritabilities of growth traits obtained from REML were 0.23-0.32 for 9 and 16 full-sib families and 0.19-0.22 for 25 full-sib families.When using Bayesian inference,the average estimated heritabilities were0.11-0.12 for 9 and 16 full-sib families and 0.13-0.16 for 25 full-sib families.Compared with REML,Bayesian got lower heritabilities,but still remained at a medium level.When the number of parents increased from 6 to 10,the estimated heritabilities were more closed to 0.20 in REML and 0.12 in Bayesian inference.Genetic correlations among traits were positive and high and had no significant difference between different sizes of designs.The accuracies of estimated breeding values from the 9 and 16 families were less precise than those from 25 families.Our results provide a basic genetic evaluation for growth traits and should be useful for the design and operation of a practical selective breeding program in the clam M.meretrix.展开更多
基金supported by National Natural Science Foundation of China(No.81403112)Beijing Natural Science Foundation(No.7154217)+1 种基金Scientific Research Program of Beijing University of Chinese Medicine(No.2015-JYB-XS104)Special Program for Beijing Key Laboratory of Chinese Medicine Manufacturing Process Control and Quality Evaluation(No.Z151100001615065)
文摘Coptis chinensis(Huanglian) is a commonly used traditional Chinese medicine(TCM) herb and alkaloids are the most important chemical constituents in it. In the present study, an isocratic reverse phase high performance liquid chromatography(RP-HPLC) method allowing the separation of six alkaloids in Huanglian was for the first time developed under the quality by design(Qb D) principles. First, five chromatographic parameters were identified to construct a Plackett-Burman experimental design. The critical resolution, analysis time, and peak width were responses modeled by multivariate linear regression. The results showed that the percentage of acetonitrile, concentration of sodium dodecyl sulfate, and concentration of potassium phosphate monobasic were statistically significant parameters(P < 0.05). Then, the Box-Behnken experimental design was applied to further evaluate the interactions between the three parameters on selected responses. Full quadratic models were built and used to establish the analytical design space. Moreover, the reliability of design space was estimated by the Bayesian posterior predictive distribution. The optimal separation was predicted at 40% acetonitrile, 1.7 g·m L-1of sodium dodecyl sulfate and 0.03 mol·m L-1 of potassium phosphate monobasic. Finally, the accuracy profile methodology was used to validate the established HPLC method. The results demonstrated that the Qb D concept could be efficiently used to develop a robust RP-HPLC analytical method for Huanglian.
基金supported by the National Natural Science Foundation of China(61104182)
文摘This study presents a Bayesian methodology for de- signing step stress accelerated degradation testing （SSADT） and its application to batteries. First, the simulation-based Bayesian de- sign framework for SSADT is presented. Then, by considering his- torical data, specific optimal objectives oriented Kullback-Leibler （KL） divergence is established. A numerical example is discussed to illustrate the design approach. It is assumed that the degrada- tion model （or process） follows a drift Brownian motion; the accele- ration model follows Arrhenius equation; and the corresponding parameters follow normal and Gamma prior distributions. Using the Markov Chain Monte Carlo （MCMC） method and WinBUGS software, the comparison shows that KL divergence is better than quadratic loss for optimal criteria. Further, the effect of simulation outiiers on the optimization plan is analyzed and the preferred sur- face fitting algorithm is chosen. At the end of the paper, a NASA lithium-ion battery dataset is used as historical information and the KL divergence oriented Bayesian design is compared with maxi- mum likelihood theory oriented locally optimal design. The results show that the proposed method can provide a much better testing plan for this engineering application.
文摘This paper discussed Bayesian variable selection methods for models from split-plot mixture designs using samples from Metropolis-Hastings within the Gibbs sampling algorithm. Bayesian variable selection is easy to implement due to the improvement in computing via MCMC sampling. We described the Bayesian methodology by introducing the Bayesian framework, and explaining Markov Chain Monte Carlo (MCMC) sampling. The Metropolis-Hastings within Gibbs sampling was used to draw dependent samples from the full conditional distributions which were explained. In mixture experiments with process variables, the response depends not only on the proportions of the mixture components but also on the effects of the process variables. In many such mixture-process variable experiments, constraints such as time or cost prohibit the selection of treatments completely at random. In these situations, restrictions on the randomisation force the level combinations of one group of factors to be fixed and the combinations of the other group of factors are run. Then a new level of the first-factor group is set and combinations of the other factors are run. We discussed the computational algorithm for the Stochastic Search Variable Selection (SSVS) in linear mixed models. We extended the computational algorithm of SSVS to fit models from split-plot mixture design by introducing the algorithm of the Stochastic Search Variable Selection for Split-plot Design (SSVS-SPD). The motivation of this extension is that we have two different levels of the experimental units, one for the whole plots and the other for subplots in the split-plot mixture design.
文摘Reliability-based design (RBD) is being adopted by geotechnical design codes worldwide, and it is therefore necessary that rock engineering practice evolves to embrace RBD. This paper examines the Hoek-Brown (H-B) strength criterion within the RBD framework, and presents three distinct analyses using a Bayesian approach. Firstly, a compilation of intact compressive strength test data for six rock types is used to examine uncertainty and variability in the estimated H-B parameters m and σc, and corresponding predicted axial strength. The results suggest that within- and between-rock type variabilities are so large that these parameters need to be determined from rock testing campaigns, rather than reference values being used. The second analysis uses an extensive set of compressive and tensile (both direct and indirect) strength data for a granodiorite, together with a new Bayesian regression model, to develop joint probability distributions of m and σc suitable for use in RBD. This analysis also shows how compressive and indirect tensile strength data may be robustly used to fit an H-B criterion. The third analysis uses the granodiorite data to investigate the important matter of developing characteristic strength criteria. Using definitions from Eurocode 7, a formal Bayesian interpretation of characteristic strength is proposed and used to analyse strength data to generate a characteristic criterion. These criteria are presented in terms of characteristic parameters mk and σck, the values of which are shown to depend on the testing regime used to obtain the strength data. The paper confirms that careful use of appropriate Bayesian statistical analysis allows the H-B criterion to be brought within the framework of RBD. It also reveals that testing guidelines such as the International Society for Rock Mechanics and Rock Engineering (ISRM) suggested methods will require modification in order to support RBD. Importantly, the need to fully understand the implications of uncertainty in nonlinear strength criteria is identified.
文摘Soil-water characteristic curve (SWCC) is significant to estimate the site-specific unsaturated soil properties (such as unsaturated shear strength and coefficient of permeability) for geotechnical analyses involving unsaturated soils. Determining SWCC can be achieved by fitting data points obtained according to the prescribed experimental scheme, which is specified by the number of measuring points and their corresponding values of the control variable. The number of measuring points is limited since direct measurement of SWCC is often costly and time-consuming. Based on the limited number of measuring points, the estimated SWCC is unavoidably associated with uncertainties, which depends on measurement data obtained from the prescribed experimental scheme. Therefore, it is essential to plan the experimental scheme so as to reduce the uncertainty in the estimated SWCC. This study presented a Bayesian approach, called OBEDO, for probabilistic experimental design optimization of measuring SWCC based on the prior knowledge and information of testing apparatus. The uncertainty in estimated SWCC is quantified and the optimal experimental scheme with the maximum expected utility is determined by Subset Simulation optimization (SSO) in candidate experimental scheme space. The proposed approach is illustrated using an experimental design example given prior knowledge and the information of testing apparatus and is verified based on a set of real loess SWCC data, which were used to generate random experimental schemes to mimic the arbitrary arrangement of measuring points during SWCC testing in practice. Results show that the arbitrary arrangement of measuring points of SWCC testing is hardly superior to the optimal scheme obtained from OBEDO in terms of the expected utility. The proposed OBEDO approach provides a rational tool to optimize the arrangement of measuring points of SWCC test so as to obtain SWCC measurement data with relatively high expected utility for uncertainty reduction.
基金This work was supported by the Natural Science Foundation of Anhui Province(2022AH050703)the National Natural Science Foundation of China(11671375).
文摘Finding the optimal dose combination in two-agent dose-finding trials is challenging due to limited sample sizes and the extensive range of potential doses.Unlike traditional chemotherapy or radiotherapy,which primarily focuses on identifying the maximum tolerated dose(MTD),therapies involving targeted and immune agents facilitate the identifica-tion of an optimal biological dose combination(OBDC)by simultaneously evaluating both toxicity and efficacy.Cur-rently,most approaches to determining the OBDC in the literature are model-based and require complex model fittings,making them cumbersome and challenging to implement.To address these challenges,we developed a novel model-as-sisted approach called uTPI-Comb.This approach refines the established utility-based toxicity probability interval design by integrating a strategically devised zone-based local and global candidate set searching strategy,which can effectively optimize the decision-making process for two-agent dose escalation or de-escalation in drug combination trials.Extensive simulation studies demonstrate that the uTPI-Comb design speeds up the dose-searching process and provides substantial improvements over existing model-based methods in determining the optimal biological dose combinations.
文摘Creating new molecules with desired properties is a fundamental and challenging problem in chemistry. Reinforcement learning (RL) has shown its utility in this area where the target chemical property values can serve as a reward signal. At each step of making a new molecule, the RL agent learns selecting an action from a list of many chemically valid actions for a given molecule, implying a great uncertainty associated with its learning. In a traditional implementation of deep RL algorithms, deterministic neural networks are typically employed, thus allowing the agent to choose one action from one sampled action at each step. In this paper, we proposed a new strategy of applying Bayesian neural networks to RL to reduce uncertainty so that the agent can choose one action from a pool of sampled actions at each step, and investigated its benefits in molecule design. Our experiments suggested the Bayesian approach could create molecules of desirable chemical quality while maintained their diversity, a very difficult goal to achieve in machine learning of molecules. We further exploited their diversity by using them to train a generative model to yield more novel drug-like molecules, which were absent in the training molecules as we know novelty is essential for drug candidate molecules. In conclusion, Bayesian approach could offer a balance between exploitation and exploration in RL, and a balance between optimization and diversity in molecule design.
基金National Natural Science Foundation of China(No.70971133)
文摘By analyzing the shortage of reliability test design and thinking over the producer's risk and consumer's risk, the information fusion technology is used to set up a reliability test design model( RTDM). By analyzing the demands and constraint conditions of the RTDM and with applications of Bayesian approach and Monte Carlo method( MCM),this paper puts forward the exponential distributed subsystems and the information fusion technology among them. According to the posteriori risk criteria,formulas of producer's risk and consumer's risk were also inferred,and with the help of Matlab software,selection of the optimum test plan was solved. Finally,validity of the model had been proved by a test of series parallel system.
基金The National High Technology Research and Development Program(863 program)of China under contract No.2012AA10A410the Zhejiang Science and Technology Project of Agricultural Breeding under contract No.2012C12907-4the Scientific and Technological Innovation Project financially supported by Qingdao National Laboratory for Marine Science and Technology under contract No.2015ASKJ02
文摘The precise and accurate knowledge of genetic parameters is a prerequisite for making efficient selection strategies in breeding programs.A number of estimators of heritability about important economic traits in many marine mollusks are available in the literature,however very few research have evaluated about the accuracy of genetic parameters estimated with different family structures.Thus,in the present study,the effect of parent sample size for estimating the precision of genetic parameters of four growth traits in clam M.meretrix by factorial designs were analyzed through restricted maximum likelihood（REML） and Bayesian.The results showed that the average estimated heritabilities of growth traits obtained from REML were 0.23-0.32 for 9 and 16 full-sib families and 0.19-0.22 for 25 full-sib families.When using Bayesian inference,the average estimated heritabilities were0.11-0.12 for 9 and 16 full-sib families and 0.13-0.16 for 25 full-sib families.Compared with REML,Bayesian got lower heritabilities,but still remained at a medium level.When the number of parents increased from 6 to 10,the estimated heritabilities were more closed to 0.20 in REML and 0.12 in Bayesian inference.Genetic correlations among traits were positive and high and had no significant difference between different sizes of designs.The accuracies of estimated breeding values from the 9 and 16 families were less precise than those from 25 families.Our results provide a basic genetic evaluation for growth traits and should be useful for the design and operation of a practical selective breeding program in the clam M.meretrix.
