Matthews Correlation Research Articles

Targeted syndromic next-generation sequencing panel for simultaneous detection of pathogens associated with bovine reproductive failure.

Bovine reproductive failure, which includes infertility, abortion, and stillbirth in cattle, leads to significant economic losses for beef and milk producers. Diagnosing the infectious causes of bovine reproductive failure is challenging as there are multiple pathogens associated with it. The traditional stepwise approach to diagnostic testing is time-consuming and can cause significant delays. In this study, we have developed a syndromic next-generation sequencing panel (BovReproSeq) for the simultaneous detection of 17 pathogens (bacteria, virus, and protozoa) associated with bovine reproductive failure. This targeted approach involves amplifying multiple pathogen-specific targets using ultra-multiplex PCR, followed by sequencing with the Oxford Nanopore platform and subsequent analysis of the data using a custom bioinformatic pipeline to determine the presence or absence of pathogens. We tested 116 clinical samples and found that BovReproSeq results matched with current diagnostic methods for 93% of the samples, and most of the disagreements occurring in samples with very low pathogen loads (Ct >35). At the optimal read-count threshold of 10 reads (minimum number of reads to classify the sample as positive), the clinical sensitivity of the assay was approximately 82%, while clinical specificity was 100%. The overall accuracy of the assay was 98.8%. Matthews correlation coefficient (correlation coefficient of binary classification) was approximately 0.90 and F1 score (harmonic mean of precision and recall) was 0.90, indicating excellent overall performance. Our study presents a significant advancement in detecting the infectious agents associated with bovine reproductive failure and the BovReproSeq panel's ability to detect 17 pathogens makes it a promising tool for veterinary diagnostics.IMPORTANCEBovine reproductive failure causes substantial economic losses to beef and milk producers, and infectious disease contributes significantly to this syndrome. Etiologic diagnosis is complicated since multiple pathogens can be involved and infections with some pathogens are asymptomatic or cause similar clinical signs. A stepwise approach to diagnostic testing is time-consuming and increases the risk of missing the correct diagnosis. BovReproSeq is a next-generation sequencing-based diagnostic panel that allows detection of 17 reproductive failure pathogens simultaneously.

Quantitative MRI Assessment of Post-Surgical Spinal Cord Injury Through Radiomic Analysis

Purpose: This study investigates radiomic efficacy in post-surgical traumatic spinal cord injury (SCI), overcoming MRI limitations from metal artifacts to enhance diagnosis, severity assessment, and lesion characterization or prognosis and therapy guidance. Background: Traumatic spinal cord injury (SCI) causes severe neurological deficits. While MRI allows qualitative injury evaluation, standard imaging alone has limitations for precise SCI diagnosis, severity stratification, and pathology characterization, which are needed to guide prognosis and therapy. Radiomics enables quantitative tissue phenotyping by extracting a high-dimensional set of descriptive texture features from medical images. However, the efficacy of postoperative radiomic quantification in the presence of metal-induced MRI artifacts from spinal instrumentation has yet to be fully explored. Methods: A total of 50 healthy controls and 12 SCI patients post-stabilization surgery underwent 3D multi-spectral MRI. Automated spinal cord segmentation was followed by radiomic feature extraction. Supervised machine learning categorized SCI versus controls, injury severity, and lesion location relative to instrumentation. Results: Radiomics differentiated SCI patients (Matthews correlation coefficient (MCC) 0.97; accuracy 1.0), categorized injury severity (MCC: 0.95; ACC: 0.98), and localized lesions (MCC: 0.85; ACC: 0.90). Combined T1 and T2 features outperformed individual modalities across tasks with gradient boosting models showing the highest efficacy. Conclusions: The radiomic framework achieved excellent performance, differentiating SCI from controls and accurately categorizing injury severity. The ability to reliably quantify SCI severity and localization could potentially inform diagnosis, prognosis, and guide therapy. Further research is warranted to validate radiomic SCI biomarkers and explore clinical integration.

PerSceptoMed 1.0: An advanced machine learning model for demographic-based antibiotic susceptibility prediction in Pseudomonas aeruginosa infections

BackgroundAntimicrobial resistance caused by Pseudomonas aeruginosa is a pressing global health issue impacting both hospitals and communities. MethodologyWe analyzed demographic and susceptibility data from 1087 samples, and tested them against five different antibiotics at the Institute of Medical Sciences & SUM Hospital. Eight machine learning algorithms (linear: multiple linear regression, generalized linear model-elastic net regularization, support vector machine-linear kernel, linear discriminant analysis; non-linear: support vector machine-radial basis function kernel, decision tree, k-nearest neighbors, kernel k-nearest neighbors), were evaluated to develop the PerSceptoMed 1.0 model for susceptibility prediction. Performance metrics such as accuracy, Matthews Correlation Coefficient, Hamming loss, 0/1 loss, and Receiver Operating Characteristic curves were computed. Statistical analysis was conducted to compare model performance. ResultsGender disparities were observed, with male patients showing higher infection rates, potentially due to occupational exposure, microbiome differences, comorbidities, and healthcare-seeking behaviors. age, clinical status, and antibiotic susceptibility were significant factors, complicating treatment strategies. Infections extended to outpatient departments, indicating a shift from hospital to community settings. Kernel k-nearest neighbors exhibited strong predictive capabilities, improving accuracy from 84.39% pre-balancing to 91.01% post-balancing, and an Area Under the Curve of 0.92. Statistical tests showed significant performance variations among models. ConclusionThe study highlights the importance of gender-sensitive approaches in infection management and the superior performance of nonlinear models in decoding complex data patterns. PerSceptoMed 1.0 demonstrates promise in enhancing antimicrobial resistance management by providing actionable insights based on demographic data.

Evaluation of Tree-Based Voting Algorithms in Water Quality Classification Prediction

Accurately predicting the state of surface water quality is crucial for ensuring the sustainable use of water resources and environmental protection. This often requires a focus on the range of factors affecting water quality, such as physical and chemical parameters. Tree models, with their flexible tree-like structure and strong capability for partitioning and selecting influential water quality features, offer clear decision-making rules, making them suitable for this task. However, an individual decision tree model has limitations and cannot fully capture the complex relationships between all influencing parameters and water quality. Therefore, this study proposes a method combining ensemble tree models with voting algorithms to predict water quality classification. This study was conducted using five surface water monitoring sites in Qingdao, representing a portion of many municipal water environment monitoring stations in China, employing a single-factor determination method with stringent surface water standards. The soft voting algorithm achieved the highest accuracy of 99.91%, and the model addressed the imbalance in original water quality categories, reaching a Matthews Correlation Coefficient (MCC) of 99.88%. In contrast, conventional machine learning algorithms, such as logistic regression and K-nearest neighbors, achieved lower accuracies of 75.90% and 91.33%, respectively. Additionally, the model’s supervision of misclassified data demonstrated its good learning of water quality determination rules. The trained model was also transferred directly to predict water quality at 13 monitoring stations in Beijing, where it performed robustly, achieving an ensemble hard voting accuracy of 97.73% and an MCC of 96.81%. In many countries’ water environment systems, different water qualities correspond to different uses, and the magnitude of influencing parameters is directly related to water quality categories; critical parameters can even directly determine the quality category. Tree models are highly capable of handling nonlinear relationships and selecting important water quality features, allowing them to identify and exploit interactions between water quality parameters, which is especially important when multiple parameters together determine the water quality category. Therefore, there is significant motivation to develop tree model-based water quality prediction models.

Prediction of Myocardial Infarction Complications using Gradient Boosting

Cardiovascular diseases (CVDs) are the leading cause of death worldwide, representing a significant public health challenge. Myocardial Infarction (MI), a severe manifestation of CVDs, contributes substantially to these fatalities. Machine learning holds great promise for predicting MI. This study explores the potential of Gradient Boosting (GB) techniques for this purpose, explicitly focusing on CatBoost, LightGBM, XGBoost, and XGBoost Random Forest. The study leverages GB's embedded feature selection, missing-value handling, and hyperparameter tuning capabilities. Performance was evaluated using multiple metrics: Area Under the Curve (AUC), classification accuracy, F1 score, precision, recall, and Matthews Correlation Coefficient (MCC). A probabilistic comparison matrix was used to assess the relative performance of the GB models. The results demonstrate the superiority of CatBoost, achieving a classification accuracy of 94.9%, an AUC of 0.992, a recall of 94.9%, and an MCC of 0.82. The probabilistic comparison further confirms CatBoost's superior performance. These findings contribute to MI prediction, highlighting the predictive potential of the CatBoost algorithm and ultimately aiding the fight against MI to achieve better patient outcomes.

Performance evaluation of the mammography ai services

Backgraund: Digital mammography is a key tool for diagnosing breast cancer (BC), which can reduce mortality by 20-40% due to early detection of pathology. To date, there have been developed many AI services that allow you to automate the process of evaluating such studies. Aims: The purpose of this work is to compare the results of different types and versions of AI services in relation to physician assessments in the analysis of mammography studies. Materials and methods: We have compared several binary scales for assessing mammography examinations, as well as several types and versions of AI services using accuracy metrics, the Matthews coefficient, and the maximum Youden index. Results: As a result of a comparative analysis, it was shown that the number of detected pathologies and the accuracy of assessing AI services depends on the binary scale for evaluating of the digital mammography. In addition, the maximum Youden index allows us to estimate the statistically significant difference between services and scales. According to this metric, AI service 1 and its version 3 have the best performance, which is consistent with most accuracy metrics. Conclusions: These results can be used in any field of medicine during the process of choosing AI services for clinical deployment.

The TabNet Model for Diagnosing Axial Spondyloarthritis Using MRI Imaging Findings and Clinical Risk Factors.

The aim of this study is to develop and validate a model for predicting axial spondyloarthritis (axSpA) based on sacroiliac joint (SIJ)-MRI imaging findings and clinical risk factors. The study is implemented on the data of 942 patients which contains of 707 patients with axSpA and 235 patients with non-axSpA. To begin with, the patients were split into training (n = 753) and validation (n = 189) cohorts. Secondly, multiple assessors manually extract the features of active inflammation (bone marrow edema) and structural lesions (erosions, sclerosis, ankylosis, joint space changes, and fat lesions). Meanwhile, we utilize 11 machine learning models and TabNet to develop imaging models, which contain six clinical risk factors for clinical models and combined clinical-imaging models. Finally, the diagnostic performance of the aforementioned models was evaluated in the validation cohort including accuracy, area under the receiver operating characteristic curve (AUC), sensitivity, specificity, F1-score, and Matthew's correlation coefficient (MCC). Six features were extracted from the imaging findings. The combined clinical-imaging models outperform the clinical and imaging models. In contrast, the combined clinical-imaging model via TabNet (CCMRT) achieved the optimal AUC of 0.93(95% CI: 0.89, 0.97). Furthermore, it is observed that the bilateral joint space changes and right-sided erosions, HLA-B27 positivity, and CRP values significantly affected axSpA diagnostic prediction. The prediction model based on clinical risk factors and SIJ-MRI imaging features can distinguish axSpA and non-axSpA effectively. In addition, the TabNet demonstrates superior diagnostic efficacy compared with machine learning models.

Groundwater Potential Mapping Using Random Forest and Extreme Gradient Boosting Algorithms

Abstract The availability of groundwater is a crucial solution to ensure the sustainability of water resources, including providing clean water. Therefore, efforts to map groundwater potential are required to enhance the efficiency of groundwater utilization and support achieving one of the Sustainable Development Goals (SDGs), particularly clean water and sanitation. This research aims to identify the distribution of the groundwater potential in Kediri Regency using the random forest (RF) and extreme gradient boosting (XGB) algorithms. This research utilizes 13 parameters, including elevation, slope, aspect, drainage density, river density, distance from rivers, lineament density, Topographic Wetness Index (TWI), Normalized Difference Vegetation Index (NDVI), land cover, soil type, lithology, and band 3 from Sentinel-2A satellite imagery. The coordinates of groundwater wells are used as training and testing data with ratios of 80:20, 70:30, and 60:40. Through the evaluation of each model’s performance using a confusion matrix, it is revealed that the best model is the RF 70:30 ratio model with Accuracy (Acc), Specificity (Spe), Sensitivity (Sen), Positive Predictive Value (PPV) values of 0.978, Cohen’s Kappa (CK) and Matthew’s Correlation Coefficient (MCC) of 0.956, and Area Under Curve (AUC) of 0.994. In this model, the elevation has the highest influence on the model, with a significance level equal to 100.

Stacking based ensemble learning framework for identification of nitrotyrosine sites

Protein nitrotyrosine is an essential post-translational modification that results from the nitration of tyrosine amino acid residues. This modification is known to be associated with the regulation and characterization of several biological functions and diseases. Therefore, accurate identification of nitrotyrosine sites plays a significant role in the elucidating progress of associated biological signs. In this regard, we reported an accurate computational tool known as iNTyro-Stack for the identification of protein nitrotyrosine sites. iNTyro-Stack is a machine-learning model based on a stacking algorithm. The base classifiers in stacking are selected based on the highest performance. The feature map employed is a linear combination of the amino composition encoding schemes, including the composition of k-spaced amino acid pairs and tri-peptide composition. The recursive feature elimination technique is used for significant feature selection. The performance of the proposed method is evaluated using k-fold cross-validation and independent testing approaches. iNTyro-Stack achieved an accuracy of 86.3% and a Matthews correlation coefficient (MCC) of 72.6% in cross-validation. Its generalization capability was further validated on an imbalanced independent test set, where it attained an accuracy of 69.32%. iNTyro-Stack outperforms existing state-of-the-art methods across both evaluation techniques. The github repository is create to reproduce the method and results of iNTyro-Stack, accessible on: https://github.com/waleed551/iNTyro-Stack/

Comparison of stable carotid plaques in patients with mild-to-moderate carotid stenosis with vulnerable plaques in patients with significant carotid stenosis.

To compare the characteristics of stable and vulnerable carotid plaques, and investigate the diagnostic performance of wall shear stress (WSS) based on magnetic resonance plaque imaging in carotid plaques. Retrospectively analyzed and divided 64 atherosclerotic plaques into stable carotid plaque groups with mild-to-moderate stenosis and vulnerable carotid plaque groups with significant stenosis. Computational fluid dynamics simulations were performed to calculate WSS parameters by using three-dimensional wall geometry based on high-resolution magnetic resonance plaque imaging of carotid bifurcation and patient specific boundary conditions obtained through color Doppler ultrasound. WSS parameters including upstream (WSSup), downstream (WSSdown), and core (WSScore) of plaque. The WSS parameters values were compared between the stable and vulnerable carotid plaque groups. Receiver operating characteristic curves and area under the curve (ROC-AUC) and Python were used to evaluate discriminative efficacy of WSS. WSSdown exhibited significant decrease in the vulnerable carotid plaque group (2.88 ± 0.41 Pa) compared to the stable carotid plaque group (4.47 ± 0.84 Pa) (P = .003). The difference of WSSup (3.28 ± 0.85 Pa vs 4.02 ± 0.74 Pa) and WSScore (1.12 ± 0.18 Pa vs 1.38 ± 0.38 Pa) between the two groups were also pronounced (P = .02, 0.01, respectively). The ROC-AUC values for WSSup, WSSdown, WSScore were 0.75 (95% CI, 0.58-0.93), 0.96 (95% CI, 0.79-1.14), 0.69 (95% CI, 0.56-0.83) respectively. When the value of WSSdown was 3.5 Pa, the sensitivity was 93.7% (95% CI, 76.1-111), specificity and accuracy was 87.5% (95% CI, 70.0-105), 88.4% (95% CI, 70.6-105) respectively. Notably, among these parameters, WSSdown demonstrated the highest discriminative efficiency with a F1 Score of 0.90, Diagnostic Odds Ratio of 105.0 and Matthews Correlation Coefficient of 0.81. Vulnerable carotid plaques with significant stenosis have lower WSS compared to stable plaques with mild-to-moderate stenosis, and downstream WSS showing the highest diagnostic efficacy.

Negative Selection Algorithm for Unsupervised Anomaly Detection

In this work, we present a modification of the well-known Negative Selection Algorithm (NSA), inspired by the process of T-cell generation in the immune system. The approach employs spherical detectors and was initially developed in the context of semi-supervised anomaly detection. The novelty of this work lies in proposing an adapted version of the NSA for unsupervised anomaly detection. The goal is to develop a method that can be applied to datasets that may not only represent self-data but also contain a small percentage of anomalies, which must be detected without prior knowledge of their locations. The proposed unsupervised algorithm leverages neighborhood sampling and ensemble methods to enhance its performance. We conducted comparative tests with 11 other algorithms across 17 datasets with varying characteristics. The results demonstrate that the proposed algorithm is competitive. The proposed algorithm performs well across multiple metrics, including accuracy, AUC, precision, recall, F1 score, Cohen’s kappa, and Matthews correlation coefficient. It consistently ranks among the top algorithms for recall, indicating its effectiveness in scenarios where detecting all existing anomalies is critical, even at the expense of some increase in false positives. Further research is possible and may focus on exploring normalization procedures, improving threshold automation, and extending the method for more detailed anomaly confidence assessments.

Ion channel classification through machine learning and protein language model embeddings.

Ion channels are critical membrane proteins that regulate ion flux across cellular membranes, influencing numerous biological functions. The resource-intensive nature of traditional wet lab experiments for ion channel identification has led to an increasing emphasis on computational techniques. This study extends our previous work on protein language models for ion channel prediction, significantly advancing the methodology and performance. We employ a comprehensive array of machine learning algorithms, including k-Nearest Neighbors, Random Forest, Support Vector Machines, and Feed-Forward Neural Networks, alongside a novel Convolutional Neural Network (CNN) approach. These methods leverage fine-tuned embeddings from ProtBERT, ProtBERT-BFD, and MembraneBERT to differentiate ion channels from non-ion channels. Our empirical findings demonstrate that TooT-BERT-CNN-C, which combines features from ProtBERT-BFD and a CNN, substantially surpasses existing benchmarks. On our original dataset, it achieves a Matthews Correlation Coefficient (MCC) of 0.8584 and an accuracy of 98.35 %. More impressively, on a newly curated, larger dataset (DS-Cv2), it attains an MCC of 0.9492 and an ROC AUC of 0.9968 on the independent test set. These results not only highlight the power of integrating protein language models with deep learning for ion channel classification but also underscore the importance of using up-to-date, comprehensive datasets in bioinformatics tasks. Our approach represents a significant advancement in computational methods for ion channel identification, with potential implications for accelerating research in ion channel biology and aiding drug discovery efforts.

Gender disparities in rural motorcycle accidents: A neural network analysis of travel behavior impact

Rural road accidents involving motorcycle riders present a formidable challenge to road safety globally. This study offers a comprehensive gender-based comparative analysis of rural road accidents among motorcycle riders, aimed at illuminating factors contributing to accidents and discerning potential gender disparities in accident rates and severity. Employing a sophisticated Neural Network approach, the research delves into the intricate relationship between various variables and accident outcomes, with a specific emphasis on identifying gender-specific patterns. For female riders, the ANN model demonstrates impressive overall accuracy (CA) of 92 %, indicating its capability to correctly classify accident outcomes. Precision, which measures the model’s ability to avoid false positives, stands at a commendable 90.8 %. Moreover, the model exhibits high recall (92 %) and F1 score (88.4 %), indicating its effectiveness in identifying both fatal and non-fatal accidents among female riders. Additionally, the Matthews Correlation Coefficient (MCC) of 0.132 suggests a moderate level of agreement between the predicted and actual outcomes. Upon further examination, it is evident that the model performs exceptionally well in predicting non-fatal accidents for female riders, achieving a precision, recall, and F1 score of 92 %, 99.9 %, and 95.8 %, respectively. However, its performance in predicting fatalities is relatively lower, with a precision of 75.6 % and recall of 2.6 %, resulting in a lower F1 score of 5.0 %. Despite this disparity, the MCC remains consistent at 0.132, indicating a balanced performance across both classes. The findings reveal valuable insights for policymakers and road safety practitioners, providing avenues for the development of targeted interventions and the enhancement of safety measures for motorcycle riders on rural roads. By addressing the gap in understanding gender-related differences in travel habits and accident risks, this research contributes to ongoing efforts to mitigate the impact of road accidents and promote safer travel environments for all road users.

A Tensor Space for Multi-View and Multitask Learning Based on Einstein and Hadamard Products: A Case Study on Vehicle Traffic Surveillance Systems

Since multi-view learning leverages complementary information from multiple feature sets to improve model performance, a tensor-based data fusion layer for neural networks, called Multi-View Data Tensor Fusion (MV-DTF), is used. It fuses M feature spaces X1,⋯,XM, referred to as views, in a new latent tensor space, S, of order P and dimension J1×⋯×JP, defined in the space of affine mappings composed of a multilinear map T:X1×⋯×XM→S—represented as the Einstein product between a (P+M)-order tensor A anda rank-one tensor, X=x(1)⊗⋯⊗x(M), where x(m)∈Xm is the m-th view—and a translation. Unfortunately, as the number of views increases, the number of parameters that determine the MV-DTF layer grows exponentially, and consequently, so does its computational complexity. To address this issue, we enforce low-rank constraints on certain subtensors of tensor A using canonical polyadic decomposition, from which M other tensors U(1),⋯,U(M), called here Hadamard factor tensors, are obtained. We found that the Einstein product A⊛MX can be approximated using a sum of R Hadamard products of M Einstein products encoded as U(m)⊛1x(m), where R is related to the decomposition rank of subtensors of A. For this relationship, the lower the rank values, the more computationally efficient the approximation. To the best of our knowledge, this relationship has not previously been reported in the literature. As a case study, we present a multitask model of vehicle traffic surveillance for occlusion detection and vehicle-size classification tasks, with a low-rank MV-DTF layer, achieving up to 92.81% and 95.10% in the normalized weighted Matthews correlation coefficient metric in individual tasks, representing a significant 6% and 7% improvement compared to the single-task single-view models.

GraphCPP: The new state-of-the-art method for cell-penetrating peptide prediction via graph neural networks.

Cell-penetrating peptides (CPPs) are short amino acid sequences that can penetrate cell membranes and deliver molecules into cells. Several models have been developed for their discovery, yet these models often face challenges in accurately predicting membrane penetration due to the complex nature of peptide-cell interactions. Hence, there is a need for innovative approaches that can enhance predictive performance. In this study, we present the application GraphCPP, a novel graph neural network (GNN) for the prediction of membrane penetration capability of peptides. A new comprehensive dataset-dubbed CPP1708-was constructed resulting in the largest reliable database of CPPs to date. Comparative analyses with previous methods, such as MLCPP2, C2Pred, CellPPD and CellPPD-Mod, demonstrated the superior predictive performance of our model. Upon testing against other published methods, GraphCPP performs exceptionally, achieving 0.5787 Matthews correlation coefficient and 0.8459 area under the curve (AUC) values on one dataset. This means a 92.8% and 23.3% improvement in Matthews correlation coefficient and AUC measures respectively compared with the next best model. The capability of the model to effectively learn peptide representations was demonstrated through t-distributed stochastic neighbour embedding plots. Additionally, the uncertainty analysis revealed that GraphCPP maintains high confidence in predictions for peptides shorter than 40 amino acids. The source code is available at https://github.com/attilaimre99/GraphCPP. These findings indicate the potential of GNN-based models to improve CPP penetration prediction and it may contribute towards the development of more efficient drug delivery systems.

Using machine learning to develop a stacking ensemble learning model for the CT radiomics classification of brain metastases.

The objective of this study was to explore the potential of machine-learning techniques in the automatic identification and classification of brain metastases from a radiomic perspective, aiming to improve the accuracy of tumor volume assessment for radiotherapy. By using various machine-learning algorithms, including random forest, support vector machine, gradient boosting machine, XGBoost, decision tree, artificial neural network, k-nearest neighbors, LightGBM, and CatBoost algorithms, a stacking ensemble model was developed to classify gross tumor volume (GTV), brainstem, and normal brain tissue based on radiomic features. Multiple evaluation metrics, including the specificity, sensitivity, negative predictive value, positive predictive value, accuracy, Matthews correlation coefficient, and the Youden index, were used to assess the model's performance. The stacked ensemble model integrated the strengths of the nine base models and consistently outperformed individual base models in classifying GTV (area under the curve [AUC] = 0.928), brainstem (AUC = 0.932), and normal brain tissue (AUC = 0.942). Among the base models, the support vector machine model demonstrated the best performance in the three classifications (AUC = 0.922, 0.909, and 0.928). The higher performance of the stacked ensemble model highlighted the low performance of other models, including the decision tree (AUC = 0.709, 0.706, 0.804) and k-nearest neighbors (AUC = 0.721, 0.663, 0.729) models in certain contexts, such as when faced with high-dimensional feature spaces. While machine learning shows significant promise in medical image analysis, relying solely on a single model may lead to suboptimal results. By combining the strengths of various algorithms, the stacking ensemble model offers a better solution for the classification of brain metastases based on radiomic features.

MCNN-AAPT: accurate classification and functional prediction of amino acid and peptide transporters in secondary active transporters using protein language models and multi-window deep learning

Secondary active transporters play a crucial role in cellular physiology by facilitating the movement of molecules across cell membranes. Identifying the functional classes of these transporters, particularly amino acid and peptide transporters, is essential for understanding their involvement in various physiological processes and disease pathways, including cancer. This study aims to develop a robust computational framework that integrates pre-trained protein language models and deep learning techniques to classify amino acid and peptide transporters within the secondary active transporter (SAT) family and predict their functional association with solute carrier (SLC) proteins. The study leverages a comprehensive dataset of 448 secondary active transporters, including 36 solute carrier proteins, obtained from UniProt and the Transporter Classification Database (TCDB). Three state-of-the-art protein language models, ProtTrans, ESM-1b, and ESM-2, are evaluated within a deep learning neural network architecture that employs a multi-window scanning technique to capture local and global sequence patterns. The ProtTrans-based feature set demonstrates exceptional performance, achieving a classification accuracy of 98.21% with 87.32% sensitivity and 99.76% specificity for distinguishing amino acid and peptide transporters from other SATs. Furthermore, the model maintains strong predictive ability for SLC proteins, with an overall accuracy of 88.89% and a Matthews Correlation Coefficient (MCC) of 0.7750. This study showcases the power of integrating pre-trained protein language models and deep learning techniques for the functional classification of secondary active transporters and the prediction of associated solute carrier proteins. The findings have significant implications for drug development, disease research, and the broader understanding of cellular transport mechanisms.

Improving the predictive capacity of the windthrow risk model ForestGALES with long-term monitoring data – A statistical calibration approach

Winter storms cause severe damage in German forests. Different modelling approaches have already been used to try and map endangered areas to minimize the risk of wind damage by stand adaption. Prevalent models for Germany include empirical-statistical and hybrid-mechanistic models, such as ForestGALES (FG). As of yet, FG is not extensively used in Germany as its parametrization requires extensive experimental efforts to derive regionally sensitive species-specific parameters. Here, we implement a statistical calibration approach for German forest conditions with observed damage from single tree data, soil types, topography (topex) and gust speed data. We use simulated annealing to generate new species-specific values for the tree species, Norway spruce, European beech, and Douglas fir from within the range of all coniferous (deciduous) species for Norway spruce and Douglas fir (European beech) and an additional 10 % buffer around the default species-specific values for each species. We compare two optimization approaches: First, we aim to maximize the Matthew’s correlation coefficient (MCC), which is calculated from the confusion matrix, applying a fixed classification threshold of 0.5. In comparison to the optimization at a fixed threshold, we optimized the species-specific parameters by maximizing the area-under-curve (AUC) value directly generated from the receiver-operator characteristic (ROC) analysis. We compare our statistical parametrizations for the considered species to those currently implemented in FG and validate the resulting damage probabilities based on confusion matrices and related performance measures. We created separate parametrizations for a single-tree and stand-wide analysis of storm damage risk, which we validated with gust speed data for Germany. Our results show, that for the single-tree method, MCC improved for all species: By 0.26 (0.22) for the calibration (validation) subset for Douglas fir, by 0.22 (0.18) for Norway spruce and by 0.08 (0.05) for European beech. The optimization for the stand-method shows an increase in MCC as well, with results not being considered due to low numbers of observation data. We show that for German forests, FG’s predictive capability can be improved by statistical optimization when no tree-pulling data is available, which could be valuable for creating further regionalizations of FG.

Prospective and External Validation of an Ensemble Learning Approach to Sensitively Detect Intravenous Fluid Contamination in Basic Metabolic Panels.

Intravenous (IV) fluid contamination within clinical specimens causes an operational burden on the laboratory when detected, and potential patient harm when undetected. Even mild contamination is often sufficient to meaningfully alter results across multiple analytes. A recently reported unsupervised learning approach was more sensitive than routine workflows, but still lacked sensitivity to mild but significant contamination. Here, we leverage ensemble learning to more sensitively detect contaminated results using an approach which is explainable and generalizable across institutions. An ensemble-based machine learning pipeline of general and fluid-specific models was trained on real-world and simulated contamination and internally and externally validated. Benchmarks for performance assessment were derived from in silico simulations, in vitro experiments, and expert review. Fluid-specific regression models estimated contamination severity. SHapley Additive exPlanation (SHAP) values were calculated to explain specimen-level predictions, and algorithmic fairness was evaluated by comparing flag rates across demographic and clinical subgroups. The sensitivities, specificities, and Matthews correlation coefficients were 0.858, 0.993, and 0.747 for the internal validation set, and 1.00, 0.980, and 0.387 for the external set. SHAP values provided plausible explanations for dextrose- and ketoacidosis-related hyperglycemia. Flag rates from the pipeline were higher than the current workflow, with improved detection of contamination events expected to exceed allowable limits for measurement error and reference change values. An accurate, generalizable, and explainable ensemble-based machine learning pipeline was developed and validated for sensitively detecting IV fluid contamination. Implementing this pipeline would help identify errors that are poorly detected by current clinical workflows and a previously described unsupervised machine learning-based method.

Classification models and SAR analysis of anaplastic lymphoma kinase (ALK) inhibitors using machine learning algorithms with two data division methods.

Anaplastic lymphoma kinase (ALK) plays a critical role in the development of various cancers. In this study, the dataset of 1810 collected inhibitors were divided into a training set and a test set by the self-organizing map (SOM) and random method, respectively. We developed 32 classification models using Support Vector Machines (SVM), Decision Trees (DT), Random Forests (RF), and Extreme Gradient Boosting (XGBoost) to distinguish between highly and weakly active ALK inhibitors, with the inhibitors represented by MACCS and ECFP4 fingerprints. Model 7D which was built by the RF algorithm using training set 1/test set 1 divided by the SOM method, provided the best performance with a prediction accuracy of 90.97% and a Matthews correlation coefficient (MCC) value of 0.79 on the test set. We clustered the 1810 inhibitors into 10 subsets by K-Means algorithm to find out the structural characteristics of highly active ALK inhibitors. The main scaffolds of highly active ALK inhibitors were also analyzed based on ECFP4 fingerprints. It was found that some substructures have a significant effect on high activity, such as 2,4-diarylaminopyrimidine analogues, pyrrolo[2,1-f][1,2,4]triazin, indolo[2,3-b]quinoline-11-one, benzo[d]imidazol and pyrrolo[2,3-b]pyridine. In addition, the subsets were summarized into several clusters, among which four clusters showed a significant relationship with ALK inhibitory activity. Finally, Shapley additive explanations (SHAP) was also used to explain the influence of modeling features on model prediction results. The SHAP results indicated that our models can well reflect the structural features of ALK inhibitors.