Changha Hwang

There is a growing interest in subject-specific predictions using neural networks, as large-scale biomedical data often exhibit dependency due to high-cardinality categorical features, which have been largely overlooked by traditional neural network frameworks. This article proposes a novel hierarchical likelihood learning framework that captures both nonlinear overall effects and subject-specific effects by incorporating gamma random effects into Poisson neural networks. The global maximizer of the proposed objective function yields maximum likelihood estimators for fixed parameters and best unbiased predictors for random effects. The proposed framework provides a robust end-to-end algorithm for clustered biomedical count data, in the sense that the corresponding estimating equations remain unbiased even when the random-effects distribution is misspecified. To enhance learning efficiency, we introduce an adjustment procedure for the random effects and variance component. Extensive simulation studies and real data analyses demonstrate the practical effectiveness of the proposed method for clustered biomedical count data. The proposed method achieves competitive predictive performance in terms of mean squared Pearson error and mean deviance across various random-effects distributions and real-world datasets.

There is a growing interest in subject-specific predictions using deep neural networks (DNNs) because real-world data often exhibit correlations, which has been typically overlooked in traditional DNN frameworks. In this paper, we propose a novel hierarchical likelihood learning framework for introducing gamma random effects into the Poisson DNN, so as to improve the prediction performance by capturing both nonlinear effects of input variables and subject-specific cluster effects. The proposed method simultaneously yields maximum likelihood estimators for fixed parameters and best unbiased predictors for random effects by optimizing a single objective function. This approach enables a fast end-to-end algorithm for handling clustered count data, which often involve high-cardinality categorical features. Furthermore, state-of-the-art network architectures can be easily implemented into the proposed h-likelihood framework. As an example, we introduce multi-head attention layer and a sparsemax function, which allows feature selection in high-dimensional settings. To enhance practical performance and learning efficiency, we present an adjustment procedure for prediction of random parameters and a method-of-moments estimator for pretraining of variance component. Various experiential studies and real data analyses confirm the advantages of our proposed methods.

Regional visit satisfaction of visitors can be increased by raising the regional attractiveness of recognized local stories. The purpose of current study is to explore the internal suitability of a confidence level for analyzing impacts to recognized local stories’ regional attractiveness empirically and the Cronbach’s alpha for the questions related to recognized local stories elements of a regional attractiveness showed good internal suitability with 0.905 and 0.874. As a result of proceeding with a regression analysis for verifying a study model, the power (the factors which are uniqueness, portrayal, understandability, and sensibility) of the recognized stories suggested from the study model and the regional attractiveness (the factors which are education, significance, and experientiality) are selected as they affect significant impacts. Local brand creations and expansions can be important cultural values for vitalizing a local economy via the study of the relationships (differences) between recognized local stories and regional attractiveness. In this study, the validity of each measure’s variable used for the study's basic static analysis and each scale construction, and the credibility whether the measures variables are consisting of latent variables consistently were verified. And it is experimented if the recognized story's regional attractiveness relationships (differences) can be applied to this study by suggesting the hypothesis adoption status via regression analysis verifying causal impacts between variables. These were very meaningful in this study. Because if we study that regional story content elements are very effective in regional activation by increasing regional attractiveness for visitors then it might be able to be systemized theoretically and be used for practical businesses.

Malware family classification is grouping malware samples that have the same or similar characteristics into the same family. It plays a crucial role in understanding notable malicious patterns and recovering from malware infections. Although many machine learning approaches have been devised for this problem, there are still several open questions including, “Which features, classifiers, and evaluation metrics are better for malware familial classification”? In this paper, we propose a machine learning approach to Android malware family classification using built-in and custom permissions. Each Android app must declare proper permissions to access restricted resources or to perform restricted actions. Permission declaration is an efficient and obfuscation-resilient feature for malware analysis. We developed a malware family classification technique using permissions and conducted extensive experiments with several classifiers on a well-known dataset, DREBIN. We then evaluated the classifiers in terms of four metrics: macrolevel F1-score, accuracy, balanced accuracy (BAC), and the Matthews correlation coefficient (MCC). BAC and the MCC are known to be appropriate for evaluating imbalanced data classification. Our experimental results showed that: (i) custom permissions had a positive impact on classification performance; (ii) even when the same classifier and the same feature information were used, there was a difference up to 3.67% between accuracy and BAC; (iii) LightGBM and AdaBoost performed better than other classifiers we considered.

Despite several improvements in the drug development pipeline over the past decade, drug failures due to unexpected adverse effects have rapidly increased at all stages of clinical trials. To improve the success rate of clinical trials, it is necessary to identify potential loser drug candidates that may fail at clinical trials. Therefore, we need to develop reliable models for predicting the outcomes of clinical trials of drug candidates, which have the potential to guide the drug discovery process. In this study, we propose an outer product–based convolutional neural network (OPCNN) model which integrates effectively chemical features of drugs and target-based features. The validation results via 10-fold cross-validations on the dataset used for a data-driven approach PrOCTOR proved that our OPCNN model performs quite well in terms of accuracy, F1-score, Matthews correlation coefficient (MCC), precision, recall, area under the curve (AUC) of the receiver operating characteristic, and area under the precision–recall curve (AUPRC). In particular, the proposed OPCNN model showed the best performance in terms of MCC, which is widely used in biomedicine as a performance metric and is a more reliable statistical measure. Through 10-fold cross-validation experiments, the accuracy of the OPCNN model is as high as 0.9758, F1 score is as high as 0.9868, the MCC reaches 0.8451, the precision is as high as 0.9889, the recall is as high as 0.9893, the AUC is as high as 0.9824, and the AUPRC is as high as 0.9979. The results proved that our OPCNN model shows significantly good prediction performance on outcomes of clinical trials and it can be quite helpful in early drug discovery.

Identifying novel drug–target interactions (DTIs) plays an important role in drug discovery. Most of the computational methods developed for predicting DTIs use binary classification, whose goal is to determine whether or not a drug–target (DT) pair interacts. However, it is more meaningful but also more challenging to predict the binding affinity that describes the strength of the interaction between a DT pair. If the binding affinity is not sufficiently large, such drug may not be useful. Therefore, the methods for predicting DT binding affinities are very valuable. The increase in novel public affinity data available in the DT-related databases enables advanced deep learning techniques to be used to predict binding affinities. In this paper, we propose a similarity-based model that applies 2-dimensional (2D) convolutional neural network (CNN) to the outer products between column vectors of two similarity matrices for the drugs and targets to predict DT binding affinities. To our best knowledge, this is the first application of 2D CNN in similarity-based DT binding affinity prediction. The validation results on multiple public datasets show that the proposed model is an effective approach for DT binding affinity prediction and can be quite helpful in drug development process.

A bstract: For a variety of reasons, most drug candidates cannot eventually pass the drug approval process. Thus, developing reliable methods for predicting clinical trial outcomes of drug candidates is crucial in improving the success rate of drug discovery and development. In this study, we propose an ensemble classifier based on weighted least squares support vector regression (LS-SVR) for predicting successes and failures of clinical trials. The efficacy of the proposed ensemble classifier is demonstrated through an experimental study on PrOCTOR dataset, which consists of informative chemical features of the drugs and target-based features. Comparing random forest and other models, the proposed ensemble classifier obtains the highest value for the area under the receiver operator curve (AUC). The results of this study demonstrate that the proposed ensemble classifier can be used to effectively predict drug approvals.

Molecular testing is increasingly important in cancer diagnosis. Targeted next generation sequencing (NGS) is widely accepted method but structural variation (SV) detection by targeted NGS remains challenging. In the brain tumor, identification of molecular alterations, including 1p/19q co-deletion, is essential for accurate glial tumor classification. Hence, we used targeted NGS to detect 1p/19q co-deletion using a newly developed deep learning (DL) model in 61 tumors, including 19 oligodendroglial tumors. An ensemble 1-dimentional convolution neural network was developed and used to detect the 1p/19q co-deletion. External validation was performed using 427 low-grade glial tumors from The Cancer Genome Atlas (TCGA). Manual review of the copy number plot from the targeted NGS identified the 1p/19q co-deletion in all 19 oligodendroglial tumors. Our DL model also perfectly detected the 1p/19q co-deletion (area under the curve, AUC = 1) in the testing set, and yielded reproducible results (AUC = 0.9652) in the validation set (n = 427), although the validation data were generated on a completely different platform (SNP Array 6.0 platform). In conclusion, targeted NGS using a cancer gene panel is a promising approach for classifying glial tumors, and DL can be successfully integrated for the SV detection in NGS data.

Spatiotemporal nonstationarity and autocorrelation are two crucial points in modeling geographical data. Previous studies have demonstrated that geographically and temporally weighted autoregressive (GTWAR) model accounts for both spatiotemporal nonstationarity and autocorrelation simultaneously to estimate house prices. Therefore, this paper proposes a kernel-based GTWAR (KBGTWAR) model by incorporating the basic principle of support vector machine regression into spatially and temporally varying coefficients model. The efficacy of KBGTWAR model is demonstrated through a case study on housing prices in the city of Shenzhen, China, from year 2004 to 2008. Comparing the existing models, KBGTWAR model obtains the lowest value for the residual sum of squares (RSS) and the highest value for the coefficient of determination R². Moreover, KBGTWAR model improves the goodness of fit of the existing GTWAR model from 12.0 to 4.5 in terms of RSS, from 0.914 to 0.968 in terms of R² and from 3.84 to 4.45 in terms of F-statistic. The results show that KBGTWAR model provides a comparatively high goodness of fit and sufficient explanatory power for both spatiotemporal nonstationarity and autocorrelation. The results of this study demonstrate that the proposed KBGTWAR model can be used to effectively formulate polices for real estate management.