Leave-one-out Cross-validation Research Articles

Radar–Rain Gauge Merging for High-Spatiotemporal-Resolution Rainfall Estimation Using Radial Basis Function Interpolation

This study introduces methods for generating fused precipitation data by applying radial basis function (RBF) interpolation, which integrates radar reflectivity-based data with ground-based precipitation gauge measurements. Rain gauges provide direct point rainfall measurements near the ground, while radars capture the spatial variability of precipitation. However, radar-based estimates, particularly for extreme rainfall events, often lack accuracy due to their indirect derivation from radar reflectivity. The study aims to produce high-resolution gridded ground precipitation data by merging radar rainfall estimates with the precise rain gauge measurements. Rain gauge data were sourced from automated synoptic observing systems (ASOSs) and automatic weather systems (AWSs), while radar data, based on hybrid surface rainfall (HSR) composites, were all provided by the Korea Meteorological Administration (KMA). Although RBF interpolation is a well-established technique, its application to the merging of radar and rain gauge data is unprecedented. To validate the accuracy of the proposed method, it was compared with traditional approaches, including the mean field bias (MFB) adjustment method, and kriging-based methods such as regression kriging (RK) and kriging with external drift (KED). Leave-one-out cross-validation (LOOCV) was performed to assess errors by analyzing overall error statistics, spatial errors, and errors in rainfall intensity data. The results showed that the RBF-based method outperformed the others in terms of accuracy.

Can overall survival (OS) benefit be predicted from improvements in progression-free survival (PFS) for previously untreated metastatic colorectal cancer (mCRC)?

222 Background: With improvements in treatment of mCRC, OS maintains its gold standard efficacy measure but takes longer to mature than intermediate endpoints. This study analyzed the association between treatment effects on PFS and OS using aggregate-level data from RCTs in previously untreated mCRC patients. Methods: A systematic literature review identified RCTs in previously untreated mCRC patients published from 2010–2021 reporting hazard ratios on PFS (HR PFS ) and OS (HR OS ). All treatments in comparison to chemotherapy alone or chemotherapy with anti-VEGF (bevacizumab) or anti-EGFR (cetuximab) targeted therapy were considered. Correlation between HR PFS and HR OS was evaluated using bivariate random-effects meta-analysis (BRMA) and weighted linear regression (WLR). Predictive performance of the surrogacy equations from WLR was assessed using leave-one-out cross-validation (LOOCV). Surrogate threshold effects (STE), defined as the minimum PFS benefit that would translate into a statistically significant OS benefit with 95% probability, were also derived to gauge the practical utility of the models. Primary analysis consisted of all included trials. Sensitivity analyses omitted trials that (I) had anti-EGFR medications, (II) had anti-VEGF medications, (III) violated proportional hazards assumptions, and (IV) permitted treatment crossover. Results: In the primary analysis 47 trials were included. The estimated correlation between PFS and OS was 0.67 (95% CI: 0.48–0.80) using BRMA and 0.70 (95% CI: 0.48–0.84) using WLR. The surrogacy equation derived from WLR was log(HR OS ) = −0.03 + 0.56 log(HR PFS ) with a statistically insignificant intercept and significant slope. Estimated STEs corresponding to sample sizes of 200 and 300 patients were 0.55 and 0.62, respectively. Observed HR OS ’s were within their 95% prediction intervals predicted from HR PFS for 93.6% of studies in LOOCV. Sensitivity analyses produced moderate correlations with >90% coverage in LOOCV (Table). Conclusions: Moderate correlations were found between HR PFS and HR OS using both modeling approaches, highlighting the stability of the findings. Cross-validations of surrogacy equations indicated promising predictive value of PFS benefit for OS benefit in previously untreated mCRC. Analysis Set​ # of Studies​ Correlation (95% CI) STE LOOCV​Coverage Rate​ BRMA WLR N = 200 N = 300 Sensitivity Analysis I​ 34​ 0.75 ​(0.57, 0.86)​ 0.78 ​(0.55, 0.90)​ 0.55​ 0.61​ 94.1​% Sensitivity Analysis II​ 14​ 0.44 ​(-0.08, 0.77)​ 0.62 ​(-0.05, 0.90)​ 0.44​ 0.52​ 92.9​% Sensitivity Analysis III​ 36​ 0.61 ​(0.36, 0.77)​ 0.59 ​(0.26, 0.80)​ 0.46​ 0.55​ 94.4​% Sensitivity Analysis IV 43​ 0.68 ​(0.48, 0.81)​ 0.67​(0.41, 0.83)​ 0.52​ 0.60​ 93.0​%

Abstract TP377: A Biomarker Based on Aneurysm Wall Enhancement and Blood Gene Expression to Identify Symptomatic Intracranial Aneurysms

Introduction: Intracranial aneurysms (IAs) are weak outpouchings on cerebral vessels that can rupture, causing subarachnoid hemorrhage. Timely and accurate risk stratification of IAs is paramount. Objective: Aneurysm wall enhancement (AWE) is a potential imaging biomarker for risk stratification. We propose to combine this with whole blood RNA sequencing to improve IA risk stratification. Methods: We retrospectively collected images and blood of patients who had undergone vessel wall imaging. We performed RNA sequencing and radiomics analysis on the IA sacs on pre- and post-contrast T1 MRI scans. Using univariate analysis, we selected significantly different radiomic features (RFs). We removed all collinear features, and pooled radiomic and expression features. We classified each IA as symptomatic or asymptomatic based on symptoms at presentation. We then built separate machine learning models: one based solely on the radiomics features and another with combined features. We performed principal component analysis (PCA) and a leave-one-out (LOO) cross validation to quantify models’ performances. Results: Our final cohort consisted of 7 patients. We found 34 significantly different RFs between symptomatic and asymptomatic IAs. The final feature set consisted of 9 RFs and 6 genes. PCA of whole dataset using the RFs alone reflected variances of 38% and 29% on the best principal components. We trained a random forest model using LOO cross validation and observed an accuracy of 57.1% (33.3% sensitivity, 75% specificity). With the addition of the gene expression features, we found the PCA explained variance to be 41% and 26%. The LOO accuracy improved to 85.7% (66.6% sensitivity, 100% specificity). Conclusion: In this study, we demonstrated that radiomics from the aneurysm wall alone have a low predictive value in identifying symptomatic IAs. However, adding gene expression levels improves the predictive value. Future work will be aimed at adding more cases.

Abstract TMP89: Predictive Model of Ischemic Event Recurrence in Pediatric Moyamoya

Background: Moyamoya is a progressive cerebral arteriopathy and one of the leading causes of stroke recurrence in children. Clinical severity scores have been developed for adult patients given the multifaceted pathophysiology of moyamoya. However, these scores were not designed to assess the risk of ischemic event recurrence, and they have not been validated in children. Objective: To develop a predictive model of ischemic event recurrence adapted to pediatric moyamoya. Methods: We retrospectively reviewed a single-center cohort of moyamoya patients, extracting demographic, clinical, and radiographic data. A binary logit mixed-effects regression was applied to predict ischemic event recurrence or disease progression based on this data. Variables were selected based on current literature. The model was validated using three approaches: validation, leave-one-out cross-validation (LOOCV), and 10-fold cross-validation, assessing performance via area under the curve (AUC), prediction accuracy, and binary classification entropy (BCE). Results: Two-hundred and twenty hemispheres were analyzed. Sixty-six hemispheres showed evidence of ischemic event recurrence or disease progression over a mean follow-up period of 6.23 ± 4.19 years. The median time to event was 21 months. Age at presentation (β=-0.036; p=0.48), Asian ancestry (β=1.74; p=0.002), asymptomatic presentation (β=1.084; p=0.207)), ischemic symptoms at presentation (β=2.15; p=0.011), evidence of infarction on initial brain MRI (β=-0.79; p=0.155), presence of an ivy sign (β=-0.2; p=0.67), unilateral disease (β=-0.97; p=0.077) with an intercept at β=-2.02 (p=0.052) were the variables included in the model. Computed model included LOOCV and 10-fold produced highest prediction accuracy (0.77 and 0.773 respectively), whereas 10-fold cross-validation had the smallest BCE value along with AUC=0.736. Conclusions: We present a pediatric-specific predictive model for the recurrence of ischemic events in moyamoya patients, demonstrating good accuracy. Our performance metrics indicate that 10-fold cross-validation is the preferred method for predicting moyamoya recurrence. This model provides a framework for developing a scoring scale that can be used in clinical practice to quantify risk and guide treatment decisions.

Serial filter-wrapper feature selection method with elite guided mutation strategy on cancer gene expression data

sNowadays, many researchers utilize cancer gene expression data to solve the problem of cancer subtype diagnosis, but cancer gene expression data are often high-dimensional, multi-sample, and multi-classified, so a hybrid serial filter-wrapper feature selection (FS) method based on elite guided mutation strategy for cancer gene expression data is proposed. It is divided into a preliminary screening phase and a combined modeling phase. In the preliminary screening stage, the threshold values of seven filter methods are determined by the leave-one cross-validation method, and the features selected by these seven filter methods are combined to form two subsets by using the thoughts of ‘‘And’’ and ‘‘Or’’ in the logical operation. The union subset of two subsets is used in the equilibrium optimizer (EO) in the subsequent combination model stage as the reserved subset in the preliminary screening stage. The resulting hybrid framework is connected by a parallel filter method designed in the first stage with an improved EO in the second stage, which is named as SFEMEO. In order to prove the effectiveness and generalization of the proposed SFEMEO, it is compared with other 9 basic algorithms on 10 UCI data sets. It is found that the classification accuracy of the SFEMEO is improved by 0.56% ~ 20.19%, and the optimal fitness is also greatly improved. After comparing SFEMEO with other nine intelligent optimization algorithms on ten cancer gene expression data sets, it can be found that compared with most algorithms, the accuracy rate is improved by 3.73% ~ 18.13%, and the optimal fitness is relatively superior. At the same time, Wilcoxon rank sum test was used to evaluate the results of intelligent optimization algorithms such as SFEMEO, which proved the effectiveness of the proposed hybrid framework and its superiority in solving the FS problem of high-dimensional cancer gene expression data.

Bundle-Specific Tractography Approach for Identifying White Matter Microstructural Changes Following Traumatic Brain Injury in Rats: An EpiBioS4Rx Study

Abstract Traumatic brain injury (TBI) presents a major global health concern, characterized by a variety of negative long term neurological outcomes. Current diagnostic tools lack the sensitivity to fully capture the complex pathophysiology of TBI and predict long-term consequences, underscoring the need for robust methods for biomarker detection. This study, conducted within the multicenter Epilepsy Bioinformatics Study for Antiepileptogenic Therapy (EpiBioS4Rx) framework, used a standardized lateral fluid-percussion injury (FPI) model to produce TBI in the left hemisphere of adult male Sprague-Dawley rats across three sites: University of Eastern Finland, Monash University, and the University of California, Los Angeles. This study utilized a novel kernel regression method for improved estimation of fiber orientations and streamline tractography to derive diffusion tensor imaging (DTI) metrics of 36 white matter tracts which were used as features to classify TBI vs of sham-operated rodents scanned at 2 days (30 sham, 87 TBI), 9 days (29 sham, 84 TBI), 1 month (28 sham, 81 TBI), and 5 months (25 sham, 65 TBI) post-injury using Elastic Net Regression regularization. A mean area under the curve (AUC) of 0.92 was achieved in correctly classifying the TBI rats in a leave-one-out cross-validation (LOOCV) framework. The results revealed delayed, region-specific effects on the microstructure of the left fimbria and left thalamic subcortical projections at 5 months following TBI. By integrating multi-compartment modeling, tractography, and harmonization, this study advances our understanding of the temporal evolution of TBI pathogenesis, paving the way for development of translational prognostic biomarkers for the risk of PTE.

Machine learning classification and biochemical characteristics in the real-time diagnosis of gastric adenocarcinoma using Raman spectroscopy

This study aimed to identify biomolecular differences between benign gastric tissues (gastritis/intestinal metaplasia) and gastric adenocarcinoma and to evaluate the diagnostic power of Raman spectroscopy-based machine learning in gastric adenocarcinoma. Raman spectroscopy-based machine learning was applied in real-time during endoscopy in 19 patients (aged 51–85 years) with high-risk for gastric adenocarcinoma. Raman spectra were captured from suspicious lesions and adjacent normal mucosa, which were biopsied for matched histopathologic diagnosis. Spectral data were analyzed using principal component analysis (PCA) and linear discriminant analysis (LDA) with leave-one-out cross-validation (LOOCV) to develop a machine learning model for diagnosing gastric adenocarcinoma. High-quality spectra (800–3300 cm⁻¹) revealed distinct patterns: adenocarcinoma tissues had higher intensities below 3150 cm⁻¹, while benign tissues exhibited higher intensities between 3150 and 3290 cm⁻¹ (p < 0.001). The model achieved diagnostic accuracy, sensitivity, specificity, and AUC values of 0.905, 0.942, 0.787, and 0.957, respectively. Biochemical correlations suggested adenocarcinoma tissues had increased protein (e.g., phenylalanine), reduced lipids, and lower water content compared to benign tissues. This study highlights the potential of Raman spectroscopy with machine learning as a real-time diagnostic tool for gastric adenocarcinoma. Further validation could establish this technique as a non-invasive, accurate method to aid clinical decision-making during endoscopy.

A Ge.F.I. Collaborative Study: Evaluating Reproducibility and Accuracy of a DNA-Methylation-Based Age-Predictive Assay for Routine Implementation in Forensic Casework.

The increasing interest in DNA methylation (DNAm) analysis within the forensic scientific community prompted a collaborative project by Ge.F.I. (Genetisti Forensi Italiani). The study evaluated a standardized bisulfite conversion-based Single Base Extension (SBE) protocol for the analysis of the methylation levels at five age-predictive loci (ELOVL2, FHL2, KLF14, C1orf132/MIR29B2C, and TRIM59). The study encompassed three phases: (1) setting up and validating the protocol to ensure consistency and reproducibility; (2) comparing fresh peripheral blood with blood spots; and (3) evaluating sources of intra- and inter-laboratory variability. Samples from 22 Italian volunteers were analyzed by 6 laboratories in replicates for a total of 528 records. From phase I emerged that the choice of genetic sequencer significantly contributed to inter-laboratory data variation, resulting in separate regression analyses performed for each laboratory. In phase II, blood spots were found to be a reliable source for DNAm analysis, despite exhibiting increased experimental variation compared to fresh peripheral blood. In phase III, a strong correlation between the individual's predicted and true ages was observed across different laboratories. Analysis of variance (ANOVA) of the residuals indicated that one-third of the total variance could be attributed to laboratory-specific factors, whereas two-thirds could be attributed to inter-individual biological differences. The leave-one-out cross-validation (LOO-CV) method yielded an overall mean absolute deviation (MAD) value of 4.41 years, with an average 95% confidence interval of 5.24 years. Stepwise regression analysis proved that a restricted model (ELOVL2, C1orf132/MIR29B2C, and TRIM59) produced results virtually indistinguishable from the five-loci model. Additionally, the analysis of samples in replicates greatly improved the fit of the regression model, balancing the slight effects of intra-laboratory variability. In conclusion, the bisulfite conversion-based SBE protocol, combined with replicate analysis and in-lab calibration of a regression-prediction model, proves to be a reliable and easily implementable method for age prediction in forensic laboratories.

Adaptive residual subsampling algorithms for kernel interpolation based on cross validation techniques

In this article we present an adaptive residual subsampling scheme designed for kernel based interpolation. For an optimal choice of the kernel shape parameter we consider some cross validation (CV) criteria, using efficient algorithms of $k$-fold CV and leave-one-out CV (LOOCV) as a special case. In this framework, the selection of the shape parameter within the residual subsampling method is totally automatic, provides highly reliable and accurate results for any kind of kernel, and guarantees existence and uniqueness of the kernel based interpolant. Numerical results show the performance of this new adaptive scheme, also giving a comparison with other computational techniques.

Radiomics for differentiation of somatic BAP1 mutation on CT scans of patients with pleural mesothelioma.

The BRCA1-associated protein 1 (BAP1) gene is of great interest because somatic (BAP1) mutations are the most common alteration associated with pleural mesothelioma (PM). Further, germline mutation of the BAP1 gene has been linked to the development of PM. This study aimed to explore the potential of radiomics on computed tomography scans to identify somatic BAP1 gene mutations and assess the feasibility of radiomics in future research in identifying germline mutations. A cohort of 149 patients with PM and known somatic BAP1 mutation status was collected, and a previously published deep learning model was used to first automatically segment the tumor, followed by radiologist modifications. Image preprocessing was performed, and texture features were extracted from the segmented tumor regions. The top features were selected and used to train 18 separate machine learning models using leave-one-out cross-validation (LOOCV). The performance of the models in distinguishing between BAP1-mutated (BAP1+) and BAP1 wild-type (BAP1-) tumors was evaluated using the receiver operating characteristic area under the curve (ROC AUC). A decision tree classifier achieved the highest overall AUC value of 0.69 (95% confidence interval: 0.60 and 0.77). The features selected most frequently through the LOOCV were all second-order (gray-level co-occurrence or gray-level size zone matrices) and were extracted from images with an applied transformation. This proof-of-concept work demonstrated the potential of radiomics to differentiate among BAP1+/- in patients with PM. Future work will extend these methods to the assessment of germline BAP1 mutation status through image analysis for improved patient prognostication.

Synchronous and anti-phase drumming elicit similar prosocial behavior ratings

PurposeMusic performance facilitates prosociality across many cultures and contexts. Interestingly, the relationship between prosociality and sensorimotor synchronization (SMS) has so far primarily been demonstrated in the context of in-phase synchrony with only a few mixed results for anti-phase coordination. In anti-phase coordination, participants move at the same rate, at opposite phases, which also requires high levels of coordination and attention. This case is particularly relevant for music and prosociality, as music regularly involves naturalistic anti-phase coordination. We thus tested whether anti-phase synchronization is as effective as in-phase synchronization at eliciting prosocial behavior.MethodsDyads (N = 50 dyads) were randomly assigned to complete four trials of a drumming sensorimotor synchronization-continuation task (SCT) either alone, synchronously or in anti-phase. Before and after the drumming task, dyads completed a behavioral economics game involving trust. Additionally, a questionnaire about trust, cooperation, affect, and similarity was given after the drumming task.ResultsCooperation rates in the stag-hunt game were near ceiling (~87%) across all conditions pre-SCT, with negligible change after the drumming task. Questionnaire items were analyzed using Bayesian probit mixed effects models to account for dyadic sampling and ordinal data, and to provide evidence in favor of the null hypothesis. Models provided moderate to extremely strong evidence that the anti-phase and in-phase coordination conditions rated their affect, trust, similarity, and cooperation more strongly than dyads in the alone condition (all BF10 &amp;gt; 3). When only comparing the anti-phase and in-phase conditions, moderate evidence in favor of the null (i.e., that phase does not affect ratings) was found for all questions (all BF10 &amp;lt; 0.3). Descriptions of the posterior, as well as leave-one-out cross validation (LOO) results, were in general accordance with the Bayes Factor results.ConclusionEvidence indicates anti-phase drumming coordination is as effective as in-phase in increasing perceived trust, cooperation, affect, and similarity. Future analyses will examine how other characteristics of the drumming coordination, such as the lag-1 autocorrelation and variability of the inter-tap interval time-series, relate to prosocial behavior and ratings of trust and cooperation.

Detecting Sensitive Spectral Bands and Vegetation Indices for Potato Yield Using Handheld Spectroradiometer Data.

Remote sensing is a valuable tool in precision agriculture due to its spatial and temporal coverage, non-destructive method of data collection, and cost-effectiveness. In this study, we measured the canopy reflectance of potato (Solanum tuberosum L.) crops on a plant-by-plant basis with a handheld spectrometer instrument. Our study pursues two primary objectives: (1) determining the optimal temporal aggregation for measuring canopy signals related to potato yield and (2) identifying the best spectral bands in the 350-2500 nm domain and vegetation indices. The study was conducted over two consecutive years (2020 and 2021) with 60 plants per plot, encompassing six potato varieties and three replicates annually throughout the growth season. Employing correlation analysis and dimensionality reduction, we identified 23 independent features significantly correlated with tuber yield. We used multiple linear regression analysis to model the relationship between the selected features and yield and to compare their influence in the fitted model. We used the Leave-One-Out Cross-Validation (LOOCV) method to assess the validity of the model (RMSE = 702 g and %RMSE = 29.2%). The most significant features included the Gitelson2 and Vogelmann indices. The optimal time period for measurements was determined to be from 56 to 100 days after planting. These findings may contribute to the advancement of precision farming by proposing tailored sensor applications, paving the way for improved agricultural practices and enhanced food security.

A motor imagery classification model based on hybrid brain-computer interface and multitask learning of electroencephalographic and electromyographic deep features.

Extracting deep features from participants' bioelectric signals and constructing models are key research directions in motor imagery (MI) classification tasks. In this study, we constructed a multimodal multitask hybrid brain-computer interface net (2M-hBCINet) based on deep features of electroencephalogram (EEG) and electromyography (EMG) to effectively accomplish motor imagery classification tasks. The model first used a variational autoencoder (VAE) network for unsupervised learning of EEG and EMG signals to extract their deep features, and subsequently applied the channel attention mechanism (CAM) to select these deep features and highlight the advantageous features and minimize the disadvantageous ones. Moreover, in this study, multitask learning (MTL) was applied to train the 2M-hBCINet model, incorporating the primary task that is the MI classification task, and auxiliary tasks including EEG reconstruction task, EMG reconstruction task, and a feature metric learning task, each with distinct loss functions to enhance the performance of each task. Finally, we designed module ablation experiments, multitask learning comparison experiments, multi-frequency band comparison experiments, and muscle fatigue experiments. Using leave-one-out cross-validation(LOOCV), the accuracy and effectiveness of each module of the 2M-hBCINet model were validated using the self-made MI-EEMG dataset and the public datasets WAY-EEG-GAL and ESEMIT. The results indicated that compared to comparative models, the 2M-hBCINet model demonstrated good performance and achieved the best results across different frequency bands and under muscle fatigue conditions. The 2M-hBCINet model constructed based on EMG and EEG data innovatively in this study demonstrated excellent performance and strong generalization in the MI classification task. As an excellent end-to-end model, 2M-hBCINet can be generalized to be used in EEG-related fields such as anomaly detection and emotion analysis.

Causal mechanisms of quadruple networks in pediatric bipolar disorder.

Pediatric bipolar disorder (PBD) is characterized by abnormal functional connectivity among distributed brain regions. Increasing evidence suggests a role for the limbic network (LN) and the triple network model in the pathophysiology of bipolar disorder (BD). However, the specific relationship between the LN and the triple network in PBD remains unclear. This study aimed to investigate the aberrant causal connections among these four core networks in PBD. Resting-state functional MRI scans from 92 PBD patients and 40 healthy controls (HCs) were analyzed. Dynamic Causal Modeling (DCM) was employed to assess effective connectivity (EC) among the four core networks. Parametric empirical Bayes (PEB) analysis was conducted to identify ECs associated with group differences, as well as depression and mania severity. Leave-one-out cross-validation (LOOCV) was used to test predictive accuracy. Compared to HCs, PBD patients exhibited primarily excitatory bottom-up connections from the LN to the salience network (SN) and bidirectional excitatory connections between the default mode network (DMN) and SN. In PBD, top-down connectivity from the triple network to the LN was excitatory in individuals with higher depression severity but inhibitory in those with higher mania severity. LOOCV identified dysconnectivity circuits involving the caudate and hippocampus as being associated with mania and depression severity, respectively. Disrupted bottom-up connections from the LN to the triple network distinguish PBD patients from healthy controls, while top-down disruptions from the triple network to LN relate to mood state differences. These findings offer insight into the neural mechanisms of PBD.

Sensory Profiling and Classification of Greek Olive Oil Varieties Using Principal Component Analysis and Learning Vector Quantization

ABSTRACTThis study aimed to explore whether the three primary sensory attributes—fruitiness, bitterness, and pungency—could be used to discriminate among seven Greek olive oil varieties. Sensory data from 110 extra virgin olive oil samples, collected over 9 years by the Kalamata Olive Oil Taste Laboratory, were analyzed using principal component analysis (PCA) and learning vector quantization (LVQ). PCA revealed that varieties with overlapping sensory profiles formed three distinct sensory groups: Koroneiki‐Lianolia‐Chalkidikis, Makri‐Manaki‐Megaritiki, and Athenolia. Consequently, the research focus shifted to evaluating the ability of these sensory attributes to differentiate between groups of varieties rather than individual varieties. LVQ, validated through leave‐one‐out cross‐validation (LOOCV), classified the olive oil samples into these sensory groups with 80% accuracy. The results demonstrate that the intensity levels of fruitiness, bitterness, and pungency provide valuable information for distinguishing groups of varieties with similar sensory profiles. While the study benefits from a robust nine‐year dataset, limitations include a relatively small sample size, potentially limiting generalizability, and a selection criterion that excludes samples with lower fruitiness scores, possibly introducing a degree of bias.Practical ApplicationsThe study's findings provide practical applications in various areas. For panel training, the identified sensory profiles can serve as benchmarks, helping tasters recognize and classify olive oil varieties. Producers can utilize these sensory profiles for consistent quality control and strategic blending, ensuring their products meet desired sensory standards. Finally, these profiles can be used in culinary education, enabling chefs and food enthusiasts to select and pair olive oils more effectively with different dishes.

Comparing Machine Learning Models for Strength and Ductility in High-Entropy Alloys

AbstractWe compare several machine learning (ML) models that predict the yield strength and plasticity of high-entropy alloys (HEAs) for achieving high-accuracy with notably low root mean square errors (RMSE). Our models, developed using a comprehensive database of single-phase body-centered cubic (BCC) HEAs and BCC + B2 HEAs (where B2 is ordered BCC), integrate advanced feature engineering reflecting the current understanding of electronic factors, atomic ordering informed by mixing enthalpy, and the D parameter associated with stacking fault energy in HEAs. This approach enables systematic comparisons of different ML models, providing deep insights into the mechanical properties of BCC and related alloys. By leveraging genetic algorithms for feature selection and meticulous hyperparameter optimization, our ML framework excels in both predictive power and interpretability. The rigorous validation process includes repeated k-fold cross-validation and leave-one-out cross-validation (LOOCV), ensuring robust generalization. Moreover, our use of Shapley Additive Explanations (SHAP) revealed critical predictors such as the testing temperature-to-melting temperature ratio $$\left(\frac{{T}_{test}}{{T}_{melt}}\right)$$ T test T melt , the mixing enthalpy $$\left(\Delta {\text{H}}_{\text{mix}}\right)$$ Δ H mix and the D parameter, which play key roles in determining plasticity and yield strength. Our work represents an advancement in designing high-performance HEAs with optimized strength and ductility, offering a powerful tool for predicting mechanical properties and exploring new candidate alloys with exceptional mechanical behavior.

Facial Movements Extracted from Video for the Kinematic Classification of Speech.

Speech Sound Disorders (SSDs) are prevalent communication problems in children that pose significant barriers to academic success and social participation. Accurate diagnosis is key to mitigating life-long impacts. We are developing a novel software solution-the Speech Movement and Acoustic Analysis Tracking (SMAAT) system to facilitate rapid and objective assessment of motor speech control issues underlying SSD. This study evaluates the feasibility of using automatically extracted three-dimensional (3D) facial measurements from single two-dimensional (2D) front-facing video cameras for classifying speech movements. Videos were recorded of 51 adults and 77 children between 3 and 4 years of age (all typically developed for age) saying 20 words from the mandibular and labial-facial levels of the Motor-Speech Hierarchy Probe Wordlist (MSH-PW). Measurements around the jaw and lips were automatically extracted from the 2D video frames using a state-of-the-art facial mesh detection and tracking algorithm, and each individual measurement was tested in a Leave-One-Out Cross-Validation (LOOCV) framework for its word classification performance. Statistics were evaluated at the α=0.05 significance level and several measurements were found to exhibit significant classification performance in both the adult and child cohorts. Importantly, measurements of depth indirectly inferred from the 2D video frames were among those found to be significant. The significant measurements were shown to match expectations of facial movements across the 20 words, demonstrating their potential applicability in supporting clinical evaluations of speech production.

Synergizing Machine Learning, Conceptual Density Functional Theory, and Biochemistry: No-Code Explainable Predictive Models for Mutagenicity in Aromatic Amines.

This study synergizes machine learning (ML) with conceptual density functional theory (CDFT) to develop OECD-compliant predictive models for the mutagenic activity of aromatic amines (AAs) with a fully No-Code methodology using a comprehensive data set of 251 AAs, Leave-One-Out-Cross-Validation (LOOCV), and three distinct data splits. Our research employs the GFN2-xTB method, known for its robustness and speed, to compute descriptors for procarcinogens and their activated metabolites in vacuum and aqueous phases. We evaluate the effectiveness of different theoretical definitions of electrophilicity within CDFT, namely, PSL, GCV, and CDP schemes, and the newly introduced Log QP descriptor to approximate Log P information. SPAARC, RandomTree, and JCHAID* ML methods were used to build explainable predictive models with highly robust internal validation (Avg. Correct Classifications = 76% and Avg. Kappa = 0.29) and external validation (Avg. Correct Classifications = 79% and Avg. Kappa = 0.33) metrics, and the results were compared to those of a two hidden layer Multilayer Perceptron. The results indicate that the second CDP definition for the electrophilicity in both vacuum and aqueous phases and also the newly presented Log QP descriptors are the most important ones for predicting the mutagenic activity of AA (namely ω+VacCDP2+, ω+AqCDP2+, and LogQP1+Vac, respectively). The results indicate that metabolic activation, aqueous solvent properties, and the CDP electrophilicity schemes and Log QP should be considered when building predictive models for the mutagenic activity of AA. This study offers a replicable, No-Code approach to QSAR research, making high-level ML and CDFT applications accessible to a broader audience. Future work will expand these methods to other compound families, enhancing predictive capabilities in the study of mutagenic activities and other biological phenomena.

Developing quantitative Adverse Outcome Pathways: An ordinary differential equation-based computational framework

Developing quantitative Adverse Outcome Pathways: An ordinary differential equation-based computational framework

Estimating monthly NO2, O3, and SO2 concentrations via an ensemble three-stage procedure with downscaled satellite remote sensing data and ground measurements

The gaseous air pollutants nitrogen dioxide (NO2), ozone (O3), and sulfur dioxide (SO2), in addition to particulate matter (PM2.5), are known to be associated with many adverse health effects. However, exposure estimates may not be available in rural or mountainous areas without monitoring stations. In this study, we retrieved satellite remote sensing data for NO2, O3, and SO2 from the Ozone Monitoring Instrument (OMI) L3 products. Together with ground measurements (air monitoring stations and meteorological and land use data), we estimated the monthly NO2, O3, and SO2 concentrations with a spatial resolution of 3×3 km across Taiwan from 2005 to 2019. A three-stage estimation procedure was utilized: in Stage 1, an ensemble generalized additive model (GAM) and machine learning method were used to determine the spatiotemporal variations; in Stage 2, the remote sensing data were downscaled; and in Stage 3, the downscaled concentrations were reused in the Stage 1 procedure for fine-tuning estimations. We obtained overall leave-one-out cross-validation (LOOCV) R2 values ranging from 0.927-0.950, 0.704-0.721, and 0.601-0.716, and root-mean-square-errors (RMSEs) ranging from 1.59-2.28, 3.81-4.18, and 0.67-1.32 ppb for NO2, O3, and SO2, respectively, via the random forest procedure. The annual NO2 and SO2 concentrations greatly improved from 2005-2019, especially in the western residential area of Taiwan. However, despite these improvements in air quality, the annual O3 concentrations tended to increase from 2015-2019. This might be due to the complex mixtures of precursors (e.g., NO2), atmospheric circulation, barriers of the Central Mountain Ridge, and increasing ground temperatures over the past decade. The proposed multistage estimation procedure performed well over the whole island with complex terrain and topography. The study outcomes may provide epidemiological information for long-term ambient exposure estimates and guidance for future administrative policies.