Supervised Statistical Learning Prediction of Soybean Varieties and Cultivation Sites Using Rapid UPLC-MS Separation, Method Validation, and Targeted Metabolomic Analysis of 31 Phenolic Compounds in the Leaves

Soybean (Glycine max; SB) leaf (SL) is an abundant non-conventional edible resource that possesses value-adding bioactive compounds. We predicted the attributes of SB based on the metabolomes of an SL using targeted metabolomics. The SB was planted in two cities, and SLs were regularly obtained from the SB plant. Nine flavonol glycosides were purified from SLs, and a validated simultaneous quantification method was used to establish rapid separation by ultrahigh-performance liquid chromatography-mass detection. Changes in 31 targeted compounds were monitored, and the compounds were discriminated by various supervised machine learning (ML) models. Isoflavones, quercetin derivatives, and flavonol derivatives were discriminators for cultivation days, varieties, and cultivation sites, respectively, using the combined criteria of supervised ML models. The neural model exhibited higher prediction power of the factors with high fitness and low misclassification rates while other models showed lower. We propose that a set of phytochemicals of SL is a useful predictor for discriminating characteristics of edible plants.

A Bi-criteria Optimization Model for Adjusting the Decision Tree Parameters

Decision trees play a very important role in knowledge representation because of its simplicity and self-explanatory nature. We study the optimization of the parameters of the decision trees to find a shorter as well as more accurate decision tree. Since these two criteria are in conflict, we need to find a decision tree with suitable parameters that can be a trade off between two criteria. Hence, we design two algorithms to build a decision tree with a given threshold of the number of vertices based on the bi-criteria optimization technique. Then, we calculate the local and global misclassification rates for these trees. Our goal is to study the effect of changing the threshold for the bi-criteria optimization of the decision trees. We apply our algorithms to 13 decision tables from UCI Machine Learning Repository and recommend the suitable threshold that can give us more accurate decision trees with a reasonable number of vertices.

Expert Knowledge Correlated Intrusion Detection System Evaluation Framework for Heterogeneous IoT

The dynamic contexts of heterogeneous Internet of Things (HetIoT) adversely affect the performance of learning-based network intrusion detection systems (NIDS) resulting in increased misclassification rates---necessitating an expert knowledge correlated evaluation framework. The proposed generalizable framework includes intrusion root cause analysis, correlation model, and correlated classification metrics that can be generalized over any NID dataset, corresponding expert knowledge, detection technique, and learning-based algorithm to facilitate context-awareness in reducing false alerts. To achieve this, we perform experimentations on the Bot-IoT dataset---with generalized traffic behaviors from multiple existing NID datasets---employing the Support Vector Machine (SVM) machine learning and Multilayer Perceptron (MLP) shallow neural network classifiers, demonstrating the generalizability, robustness, and improved performance of the propounded framework compared to the existing literature. Furthermore, the proposed framework offers minimal processing overhead on the classifier algorithms.<br>

Tri-Ponderal Mass Index Reference Values for Screening Metabolic Syndrome in Children and Adolescents: Results From Two National-Representative Cross-Sectional Studies in China and America

IntroductionTo ascertain the possible cut point of tri-ponderal mass index (TMI) in discriminating metabolic syndrome (MetS) and related cardio-metabolic risk factors in Chinese and American children and adolescents.MethodsA total of 57,201 Chinese children aged 7-18 recruited in 2012 and and 10,441 American children aged 12-18 from National Health and Nutrition Examination Survey (NHANES 2001-2014) were included to fit TMI percentiles. Participants were randomly assigned to a derivation set (75%) and validation set (25%). The cut points of TMI with the lowest misclassification rate under the premise of the highest area under curves (AUC) were selected for each sex, which were additionally examined in the validation set. All of data analysis was conducted between September and December in 2019.ResultsTMI showed good capacity on discriminating MetS, with AUC of 0.7658 (95% CI: 0.7544-0.7770) to 0.8445 (95% CI: 0.8349-0.8537) in Chinese and 0.8871 (95% CI: 0.8663-0.9056) to 0.9329 (95% CI: 0.9166-0.9469) in American children. The optimal cut points were 14.46 kg/m3 and 13.91 kg/m3 for Chinese boys and girls, and 17.08 kg/m3 and 18.89 kg/m3 for American boys and girls, respectively. The corresponding misclassification rates were 17.1% (95% CI: 16.4-17.8) and 11.2% (95% CI: 9.9-12.6), respectively. Performance of these cut points were also examined in the validation set (sensitivity 67.7%, specificity 82.4% in Chinese; sensitivity 84.4%, specificity 88.7% in American children).ConclusionsA sex- and ethnicity- specific single cut point of TMI could be used to distinguish MetS and elevated risk of cardio-metabolic factors in children and adolescents.

A Comparison of Different Approaches to Clinical Phenotyping of Lithium Response: A Proof of Principle Study Employing Genetic Variants of Three Candidate Circadian Genes

Optimal classification of the response to lithium (Li) is crucial in genetic and biomarker research. This proof of concept study aims at exploring whether different approaches to phenotyping the response to Li may influence the likelihood of detecting associations between the response and genetic markers. We operationalized Li response phenotypes using the Retrospective Assessment of Response to Lithium Scale (i.e., the Alda scale) in a sample of 164 cases with bipolar disorder (BD). Three phenotypes were defined using the established approaches, whilst two phenotypes were generated by machine learning algorithms. We examined whether these five different Li response phenotypes showed different levels of statistically significant associations with polymorphisms of three candidate circadian genes (RORA, TIMELESS and PPARGC1A), which were selected for this study because they were plausibly linked with the response to Li. The three original and two revised Alda ratings showed low levels of discordance (misclassification rates: 8–12%). However, the significance of associations with circadian genes differed when examining previously recommended categorical and continuous phenotypes versus machine-learning derived phenotypes. Findings using machine learning approaches identified more putative signals of the Li response. Established approaches to Li response phenotyping are easy to use but may lead to a significant loss of data (excluding partial responders) due to recent attempts to improve the reliability of the original rating system. While machine learning approaches require additional modeling to generate Li response phenotypes, they may offer a more nuanced approach, which, in turn, would enhance the probability of identifying significant signals in genetic studies.

Expert Knowledge Correlated Evaluation of Intrusion Detection System in Heterogeneous IoT

<div>The dynamic heterogeneous IoT contexts adversely affect the performance of learning-based network intrusion detection and prevention systems resulting in increased misclassification rates—necessitating an expert knowledge correlated evaluation framework. The proposed framework includes intrusion root cause analysis and a correlation model that can be generalized over any network intrusion dataset, corresponding expert knowledge, detection technique, and learning-based algorithm. The experimentations prove the robustness of the propounded</div><div>framework on imbalanced datasets.</div>

Expert Knowledge Correlated Evaluation of Intrusion Detection System in Heterogeneous IoT

<div>The dynamic heterogeneous IoT contexts adversely affect the performance of learning-based network intrusion detection and prevention systems resulting in increased misclassification rates—necessitating an expert knowledge correlated evaluation framework. The proposed framework includes intrusion root cause analysis and a correlation model that can be generalized over any network intrusion dataset, corresponding expert knowledge, detection technique, and learning-based algorithm. The experimentations prove the robustness of the propounded</div><div>framework on imbalanced datasets.</div>

Use of 816 consecutive burn wound biopsies to inform a histologic algorithm for burn depth categorization

Introduction Burn experts are only 77% accurate when subjectively assessing burn depth, leaving almost a quarter of patients to undergo unnecessary surgery or conversely suffer a delay in treatment. To aid clinicians in burn depth assessment (BDA), new technologies are being studied with machine learning algorithms calibrated to histologic standards. Our group has iteratively created a theoretical burn biopsy algorithm (BBA) based on histologic analysis, and subsequently informed it with the largest burn wound biopsy repository in the literature. Here, we sought to report that process. Methods The was an IRB-approved, prospective, multicenter study. A BBA was created a priori and refined in an iterative manner. Patients with burn wounds assessed by burn experts as requiring excision and autograft underwent 4mm biopsies procured every 25cm 2. Serial still photos were obtained at enrollment and at excision intraoperatively. Burn biopsies were histologically assessed for presence/absence of epidermis, papillary dermis, reticular dermis, and proportion of necrotic adnexal structures by a dermatopathologist using H&E with whole slide scanning. First degree and superficial 2 nd degree were considered to be burn wounds likely to have healed without surgery, while deep 2 nd and 3 rd degree burns were considered unlikely to heal by 21 days. Biopsy pathology results were correlated with still photos by five burn experts for consensus of final burn depth diagnosis. Results Sixty-six subjects were enrolled with 117 wounds and 816 biopsies. The BBA was used to categorize subjects’ wounds into 4 categories: 7% of burns were categorized as 1 st degree, 13% as superficial 2 nd degree, 43% as deep 2 nd degree, and 37% as 3 rd degree. Therefore 20% of burn wounds were incorrectly judged as needing excision and grafting by the clinical team as per the BBA. As H&E is unable to assess the viability of papillary and reticular dermis, with time our team came to appreciate the greater importance of adnexal structure necrosis over dermal appearance in assessing healing potential. Conclusions Our study demonstrates that a BBA with objective histologic criteria can be used to categorize BDA with clinical misclassification rates consistent with past literature. This study serves as the largest analysis of burn biopsies by modern day burn experts and the first to define histologic parameters for BDA.

Estimating Probabilities of Passing for Examinees With Incomplete Data in Mastery Tests

Administrative problems such as computer malfunction and power outage occasionally lead to missing item scores and hence to incomplete data on mastery tests such as the AP and U.S. Medical Licensing examinations. Investigators are often interested in estimating the probabilities of passing of the examinees with incomplete data on mastery tests. However, there is a lack of research on this estimation problem. The goal of this article is to suggest two new approaches—one each based on classical test theory and item response theory—for estimating the probabilities of passing of the examinees with incomplete data on mastery tests. The two approaches are demonstrated to have high accuracy and negligible misclassification rates.

Investigating the Impact of Noneffortful Responses on Individual-Level Scores: Can the Effort-Moderated IRT Model Serve as a Solution?

Suboptimal effort is a major threat to valid score-based inferences. While the effects of such behavior have been frequently examined in the context of mean group comparisons, minimal research has considered its effects on individual score use (e.g., identifying students for remediation). Focusing on the latter context, this study addressed two related questions via simulation and applied analyses. First, we investigated how much including noneffortful responses in scoring using a three-parameter logistic (3PL) model affects person parameter recovery and classification accuracy for noneffortful responders. Second, we explored whether improvements in these individual-level inferences were observed when employing the Effort Moderated IRT (EM-IRT) model under conditions in which its assumptions were met and violated. Results demonstrated that including 10% noneffortful responses in scoring led to average bias in ability estimates and misclassification rates by as much as 0.15 SDs and 7% respectively. These results were mitigated when employing the EM-IRT model, particularly when model assumptions were met. However, once model assumptions were violated, the EM-IRT model’s performance deteriorated, though still outperforming the 3PL model. Thus, findings from this study show that: (a) including noneffortful responses when using individual scores can lead to potential unfounded inferences and potential score misuse; and (b) the negative impact that noneffortful responding has on person ability estimates and classification accuracy can be mitigated by employing the EM-IRT model, particularly when its assumptions are met.