Thanakorn Naenna

A practical overview of quantitative structure-activity relationship

Quantitative structure-activity relationship (QSAR) modeling pertains to the construction of predictive models of biological activities as a function of structural and molecular information of a compound library. The concept of QSAR has typically been used for drug discovery and development and has gained wide applicability for correlating molecular information with not only biological activities but also with other physicochemical properties, which has therefore been termed quantitative structure-property relationship (QSPR). Typical molecular parameters that are used to account for electronic properties, hydrophobicity, steric effects, and topology can be determined empirically through experimentation or theoretically via computational chemistry. A given compilation of data sets is then subjected to data pre-processing and data modeling through the use of statistical and/or machine learning techniques. This review aims to cover the essential concepts and techniques that are relevant for performing QSAR/QSPR studies through the use of selected examples from our previous work.

Prediction of GFP Spectral Properties Using Artificial Neural Network

The prediction of the excitation and the emission maxima of green fluorescent protein (GFP) chromophores were investigated by a quantitative structure‐property relationship study. A data set of 19 GFP color variants and an additional data set consisting of 29 synthetic GFP chromophores were collected from the literature. Artificial neural network implementing the back‐propagation algorithm was employed. The proposed computational approach reliably predicted the excitation and the emission maxima of GFP chromophores with correlation coefficient exceeding 0.9. The usefulness of quantum chemical descriptors was revealed by a comparative study with other molecular descriptors. Assignment of appropriate protonation state of the chromophore for the GFP color variants data set was shown to be necessary for good predictive performance. Results suggest that the confinement of the GFP chromophore has no significant influence on the predictive performance of the data set used. A comparative investigation with the traditional modeling methods, particularly multiple linear regression and partial least squares, reveals that artificial neural network is the most suitable modeling approach for the GFP spectral properties. It is anticipated that this methodology has great potential in accelerating the design and engineering of novel GFP color variants of scientific or industrial interest.

Quantitative prediction of imprinting factor of molecularly imprinted polymers by artificial neural network.

SummaryArtificial neural network (ANN) implementing the back-propagation algorithm was applied for the calculation of the imprinting factors (IF) of molecularly imprinted polymers (MIP) as a function of the computed molecular descriptors of template and functional monomer molecules and mobile phase descriptors. The dataset used in our study were obtained from the literature and classified into two distinctive datasets on the basis of the polymer’s morphology, irregularly sized MIP and uniformly sized MIP datasets. Results revealed that artificial neural network was able to perform well on datasets derived from uniformly sized MIP (n=23, r=0.946, RMS=2.944) while performing poorly on datasets derived from irregularly sized MIP (n=75, r=0.382, RMS=6.123). The superior performance of the uniformly sized MIP dataset over the irregularly sized MIP dataset could be attributed to its more predictable nature owing to the consistency of MIP particles, uniform number and association constant of binding sites, and minimal deviation of the imprinted polymers. The ability to predict the imprinting factor of imprinted polymer prior to performing actual experimental work provide great insights on the feasibility of the interaction between template-functional monomer pairs.

Prediction of bond dissociation enthalpy of antioxidant phenols by support vector machine

Antioxidants play crucial roles in scavenging oxidative damages arising from reactive oxygen species. Bond dissociation enthalpy (BDE) of phenolic O-H bond has well been accepted as an indicator of antioxidant activity since phenols donate the hydrogen atom to the free radicals thereby neutralizing its toxic effect. The BDEs from a data set of 39 antioxidant phenols were modeled using computationally inexpensive quantum chemical descriptors with multiple linear regression (MLR), partial least squares (PLS), and support vector machine (SVM). The molecular descriptors of the phenols were derived from calculations at the following theoretical levels: AM1, HF/3-21g(d), B3LYP/3-21g(d), and B3LYP/6-31g(d). Results indicated that when MLR and PLS were used as the regression methods, B3LYP/3-21g(d) gave the best performance with leave-one-out cross-validated correlation coefficients (r) of 0.917 and 0.921, respectively, while the semiempirical AM1 provided slightly lower r of 0.897 and 0.888, respectively. When SVM was used as the regression method no significant difference in the accuracy was observed for models using B3LYP/3-21g(d) and AM1 as indicated by r of 0.968 and 0.966, respectively. The quantitative structure-property relationship (QSPR) model of BDE discussed in this study offers great potential for the design of novel antioxidant phenols with robust properties.

Use of machine learning for classification of magnetocardiograms

We describe the use of machine learning for pattern recognition in magnetocardiography (MCG) that measures magnetic fields emitted by the electrophysiological activity of the heart. We used direct kernel methods to separate abnormal MCG heart patterns from normal ones. For unsupervised learning, we introduced Direct Kernel based Self-Organizing Maps. For supervised learning we used Direct Kernel Partial Least Squares and (Direct) Kernel Ridge Regression. These results are then compared with classical Support Vector Machines and Kernel Partial Least Squares. The hyperparameters for these methods were tuned on a validation subset of the training data before testing. We also investigated the most effective pre-processing, using local, vertical, horizontal and two-dimensional (global) Mahanalobis scaling, wavelet transforms and experimented with variable selection by filtering. The results, similar for all three methods, were encouraging, exceeding the quality of classification achieved by the trained experts.

Modeling the activity of furin inhibitors using artificial neural network.

Quantitative structure-activity relationship (QSAR) models were constructed for predicting the inhibition of furin-dependent processing of anthrax protective antigen of substituted guanidinylated aryl 2,5-dideoxystreptamines. Molecular descriptors calculated by E-Dragon and RECON were subjected to variable reduction using the Unsupervised Forward Selection (UFS) algorithm. The variables were then used as input for QSAR model generation using partial least squares and back-propagation neural network. Prediction was performed via a two-step approach: (i) perform classification to determine whether the molecule is active or inactive, (ii) develop a QSAR regression model of active molecules. Both classification and regression models yielded good results with RECON providing higher accuracy than that of E-DRAGON descriptors. The performance of the regression model using E-Dragon and RECON descriptors provided a correlation coefficient of 0.807 and 0.923 and root mean square error of 0.666 and 0.304, respectively. Interestingly, it was observed that appropriate representations of the protonation states of the molecules were crucial for good prediction performance, which coincides with the fact that the inhibitors interact with furin via electrostatic forces. The results provide good prospect of using the proposed QSAR models for the rational design of novel therapeutic furin inhibitors toward anthrax and furin-dependent diseases.

Modeling the LPS Neutralization Activity of Anti-Endotoxins

Bacterial lipopolysaccharides (LPS), also known as endotoxins, are major structural components of the outer membrane of Gram-negative bacteria that serve as a barrier and protective shield between them and their surrounding environment. LPS is considered to be a major virulence factor as it strongly stimulates the secretion of pro-inflammatory cytokines which mediate the host immune response and culminating in septic shock. Quantitative structure-activity relationship studies of the LPS neutralization activities of anti-endotoxins were performed using charge and quantum chemical descriptors. Artificial neural network implementing the back-propagation algorithm was selected for the multivariate analysis. The predicted activities from leave-one-out cross-validation were well correlated with the experimental values as observed from the correlation coefficient and root mean square error of 0.930 and 0.162, respectively. Similarly, the external testing set also yielded good predictivity with correlation coefficient and root mean square error of 0.983 and 0.130. The model holds great potential for the rational design of novel and robust compounds with enhanced neutralization activity.

Fuzzy FMEA application to improve decision-making process in an emergency department

Fuzzy logic approach is preferable to fix the drawbacks for reprioritization of the Risk Priority Number (RPN).Fuzzy logic could reduce the drawback of occurred Traditional FMEA in evaluation and prioritization of failures.The application of using Fuzzy FMEA in the emergency department can be adopted suitably. All of members were able to assess dependently without any bias from the team members.Fuzzy FMEA can be applied for the first time to improve decision making process in an emergency department of a public hospital. Hospitals are one of the important service industries of health care for patients. The emergency department is the heart of every hospital, because the errors or failures occurring in it will significantly affect the safety of patients and the goodwill of the hospital. Therefore, emergency departments should be monitored carefully. This study proposed the application of Fuzzy failure mode and effects analysis (FMEA) for prioritization and assessment of failures that likely occur in the working process of an emergency department. All individuals were assessed independently without the interference of team members. In addition, this method could reduce the limitations of traditional FMEA. The prioritization of risks could also help the emergency department to choose corrective actions wisely. In conclusion, the Fuzzy FMEA method was found to be suitably adopted in the emergency department. Finally, this method helped to increase the level of confidence on hospitals.

QSAR MODEL OF THE QUORUM-QUENCHING N-ACYL-HOMOSERINE LACTONE LACTONASE ACTIVITY

A quantitative structure-activity relationship (QSAR) study was performed to model the lactonolysis activity of N-acyl-homoserine lactone lactonase. A data set comprising of 20 homoserine lactones and related compounds was taken from the work of Wang et al. Quantum chemical descriptors were calculated using the semiempirical AM1 method. Partial least squares regression was utilized to construct a predictive model. This computational approach reliably reproduced the lactonolysis activity with high accuracy as illustrated by the correlation coefficient in excess of 0.9. It is demonstrated that the combined use of quantum chemical descriptors with partial least squares regression are suitable for modeling the AHL lactonolysis activity.

Identification of ischemic heart disease via machine learning analysis on magnetocardiograms

Ischemic heart disease (IHD) is predominantly the leading cause of death worldwide. Early detection of IHD may effectively prevent severity and reduce mortality rate. Recently, magnetocardiography (MCG) has been developed for the detection of heart malfunction. Although MCG is capable of monitoring the abnormal patterns of magnetic field as emitted by physiologically defective heart, data interpretation is time-consuming and requires highly trained professional. Hence, we propose an automatic method for the interpretation of IHD pattern of MCG recordings using machine learning approaches. Two types of machine learning techniques, namely back-propagation neural network (BNN) and direct kernel self-organizing map (DK-SOM), were applied to explore the IHD pattern recorded by MCG. Data sets were obtained by sequential measurement of magnetic field emitted by cardiac muscle of 125 individuals. Data were divided into training set and testing set of 74 cases and 51 cases, respectively. Predictive performance was obtained by both machine learning approaches. The BNN exhibited sensitivity of 89.7%, specificity of 54.5% and accuracy of 74.5%, while the DK-SOM provided relatively higher prediction performance with a sensitivity, specificity and accuracy of 86.2%, 72.7% and 80.4%, respectively. This finding suggests a high potential of applying machine learning approaches for high-throughput detection of IHD from MCG data.