A Novel Method for Identification of Glutarylation Sites Combining Borderline-SMOTE With Tomek Links Technique in Imbalanced Data

A review of machine learning methods for imbalanced data challenges in chemistry

Imbalanced data, where certain classes are significantly underrepresented in a dataset, is a widespread machine learning (ML) challenge across various fields of chemistry, yet it remains inadequately addressed. This data imbalance can lead to biased ML or deep learning (DL) models, which fail to accurately predict the underrepresented classes, thus limiting the robustness and applicability of these models. With the rapid advancement of ML and DL algorithms, several promising solutions to this issue have emerged, prompting the need for a comprehensive review of current methodologies. In this review, we examine the prominent ML approaches used to tackle the imbalanced data challenge in different areas of chemistry, including resampling techniques, data augmentation techniques, algorithmic approaches, and feature engineering strategies. Each of these methods is evaluated in the context of its application across various aspects of chemistry, such as drug discovery, materials science, cheminformatics, and catalysis. We also explore future directions for overcoming the imbalanced data challenge and emphasize data augmentation via physical models, large language models (LLMs), and advanced mathematics. The benefit of balanced data in new material design and production and the persistent challenges are discussed. Overall, this review aims to elucidate the prevalent ML techniques applied to mitigate the impacts of imbalanced data within the field of chemistry and offer insights into future directions for research and application.

Addressing data imbalance in collision risk prediction with active generative oversampling

Data imbalance is a critical factor affecting the predictive accuracy in collision risk assessment. This study proposes an advanced active generative oversampling method based on Query by Committee (QBC) and Auxiliary Classifier Generative Adversarial Network (ACGAN), integrated with the Wasserstein Generative Adversarial Network (WGAN) framework. Our method selectively enriches minority class samples through QBC and diversity metrics to enhance the diversity of sample generation, thereby improving the performance of fault classification algorithms. By equating the labels of selected samples to those of real samples, we increase the accuracy of the discriminator, forcing the generator to produce more diverse outputs, which is expected to improve classification results. We also propose a method for dynamically adjusting the training epochs of the generator and discriminator based on loss differences to achieve balance in model training. Empirical analysis on four publicly available imbalanced datasets shows that our method outperforms existing methods in terms of precision, recall, F-measure, and G-mean. Specifically, our method's results are above 0.92 on all evaluation indicators, with an average improvement of 23-28.3% compared to the worst-performing ENN method. This indicates that our method has a significant advantage in handling data imbalance, being able to more accurately identify collision samples and reduce the misclassification rate of non-collision samples.

DeepPhoPred: Accurate Deep Learning Model to Predict Microbial Phosphorylation

Phosphorylation is a substantial posttranslational modification of proteins that refers to adding a phosphate group to the amino acid side chain after translation process in the ribosome. It is vital to coordinate cellular functions, such as regulating metabolism, proliferation, apoptosis, subcellular trafficking, and other crucial physiological processes. Phosphorylation prediction in a microbial organism can assist in understanding pathogenesis and host-pathogen interaction, drug and antibody design, and antimicrobial agent development. Experimental methods for predicting phosphorylation sites are costly, slow, and tedious. Hence low-cost and high-speed computational approaches are highly desirable. This paper presents a new deep learning tool called DeepPhoPred for predicting microbial phospho-serine (pS), phospho-threonine (pT), and phospho-tyrosine (pY) sites. DeepPhoPred incorporates a two-headed convolutional neural network architecture with the squeeze and excitation blocks followed by fully connected layers that jointly learn significant features from the peptide's structural and evolutionary information to predict phosphorylation sites. Our empirical results demonstrate that DeepPhoPred significantly outperforms the existing microbial phosphorylation site predictors with its highly efficient deep-learning architecture. DeepPhoPred as a standalone predictor, all its source codes, and our employed datasets are publicly available at https://github.com/faisalahm3d/DeepPhoPred.

Compact Class-Conditional Attribute Category Clustering: Amino Acid Grouping for Enhanced HIV-1 Protease Cleavage Classification

Categorical attributes are common in many classification tasks, presenting certain challenges as the number of categories grows. This situation can affect data handling, negatively impacting the building time of models, their complexity and, ultimately, their classification performance. In order to mitigate these issues, this research proposes a novel preprocessing technique for grouping attribute categories in classification datasets. This approach combines the exact representation of the association between categorical values in a Euclidean space, clustering methods and attribute quality metrics to group similar attribute categories based on their contribution to the classification task. To estimate its effectiveness, the proposal is evaluated within the context of HIV-1 protease cleavage site prediction, where each attribute represents an amino acid that can take multiple possible values. The results obtained on HIV-1 real-world datasets show a significant reduction in the number of categories per attribute, with an average reduction percentage ranging from 74% to 81%. This reduction leads to simplified data representations and improved classification performances compared to not preprocessing. Specifically, improvements of up to 0.07 in accuracy and 0.19 in geometric mean are observed across different datasets and classification algorithms. Additionally, extensive simulations on synthetic datasets with varied characteristics are carried out, providing consistent and reliable results that validate the robustness of the proposal. These findings highlight the capability of the developed method to enhance cleavage prediction, which could potentially contribute to understanding viral processes and developing targeted therapeutic strategies.

Multinomial classification of NLRP3 inhibitory compounds based on large scale machine learning approaches

The role of NLRP3 inflammasome in innate immunity is newly recognized. The NLRP3 protein is a family of nucleotide-binding and oligomerization domain-like receptors as well as a pyrin domain-containing protein. It has been shown that NLRP3 may contribute to the development and progression of a variety of diseases, such as multiple sclerosis, metabolic disorders, inflammatory bowel disease, and other auto-immune and auto-inflammatory conditions. The use of machine learning methods in pharmaceutical research has been widespread for several decades. An important objective of this study is to apply machine learning approaches for the multinomial classification of NLRP3 inhibitors. However, data imbalances can affect machine learning. Therefore, a synthetic minority oversampling technique (SMOTE) has been developed to increase the sensitivity of classifiers to minority groups. The QSAR modelling was performed using 154 molecules retrieved from the ChEMBL database (version 29). The accuracy of the multiclass classification top six models was found to fall within ranges of 0.99 to 0.86, and log loss ranges of 0.2 to 2.3, respectively. The results showed that the receiver operating characteristic curve (ROC) plot values significantly improved when tuning parameters were adjusted and imbalanced data was handled. Moreover, the results demonstrated that SMOTE offers a significant advantage in handling imbalanced datasets as well as substantial improvements in overall accuracy of machine learning models. The top models were then used to predict data from unseen datasets. In summary, these QSAR classification models exhibited robust statistical results and were interpretable, which strongly supported their use for rapid screening of NLRP3 inhibitors.

Prediction of blood-brain barrier penetrating peptides based on data augmentation with Augur

BACKGROUND

The blood-brain barrier serves as a critical interface between the bloodstream and brain tissue, mainly composed of pericytes, neurons, endothelial cells, and tightly connected basal membranes. It plays a pivotal role in safeguarding brain from harmful substances, thus protecting the integrity of the nervous system and preserving overall brain homeostasis. However, this remarkable selective transmission also poses a formidable challenge in the realm of central nervous system diseases treatment, hindering the delivery of large-molecule drugs into the brain. In response to this challenge, many researchers have devoted themselves to developing drug delivery systems capable of breaching the blood-brain barrier. Among these, blood-brain barrier penetrating peptides have emerged as promising candidates. These peptides had the advantages of high biosafety, ease of synthesis, and exceptional penetration efficiency, making them an effective drug delivery solution. While previous studies have developed a few prediction models for blood-brain barrier penetrating peptides, their performance has often been hampered by issue of limited positive data.

RESULTS

In this study, we present Augur, a novel prediction model using borderline-SMOTE-based data augmentation and machine learning. we extract highly interpretable physicochemical properties of blood-brain barrier penetrating peptides while solving the issues of small sample size and imbalance of positive and negative samples. Experimental results demonstrate the superior prediction performance of Augur with an AUC value of 0.932 on the training set and 0.931 on the independent test set.

CONCLUSIONS

This newly developed Augur model demonstrates superior performance in predicting blood-brain barrier penetrating peptides, offering valuable insights for drug development targeting neurological disorders. This breakthrough may enhance the efficiency of peptide-based drug discovery and pave the way for innovative treatment strategies for central nervous system diseases.

iMRSAPred: Improved Prediction of Anti-MRSA Peptides Using Physicochemical and Pairwise Contact-Energy Properties of Amino Acids

Methicillin-resistant Staphylococcus aureus (MRSA) is a growing concern for human lives worldwide. Anti-MRSA peptides act as potential antibiotic agents and play significant role to combat MRSA infection. Traditional laboratory-based methods for annotating Anti-MRSA peptides are although precise but quite challenging, costly, and time-consuming. Therefore, computational methods capable of identifying Anti-MRSA peptides accelerate the drug designing process for treating bacterial infections. In this study, we developed a novel sequence-based predictor "iMRSAPred" for screening Anti-MRSA peptides by incorporating energy estimation and physiochemical and sequential information. We successfully resolved the skewed imbalance phenomena by using synthetic minority oversampling technique plus Tomek link (SMOTETomek) algorithm. Furthermore, the Shapley additive explanation method was leveraged to analyze the impact of top-ranked features in the prediction task. We evaluated multiple machine learning algorithms, i.e., CatBoost, Cascade Deep Forest, Kernel and Tree Boosting, support vector machine, and HistGBoost classifiers by 10-fold cross-validation and independent testing. The proposed iMRSAPred method significantly improved the overall performance in terms of accuracy and Matthew's correlation coefficient (MCC) by 5.45 and 0.083%, respectively, on the training data set. On the independent data set, iMRSAPred improved accuracy and MCC by 3.98 and 0.055%, respectively. We believe that the proposed method would be useful in large-scale Anti-MRSA peptide prediction and provide insights into other bioactive peptides.

GBDT_KgluSite: An improved computational prediction model for lysine glutarylation sites based on feature fusion and GBDT classifier

BACKGROUND

Lysine glutarylation (Kglu) is one of the most important Post-translational modifications (PTMs), which plays significant roles in various cellular functions, including metabolism, mitochondrial processes, and translation. Therefore, accurate identification of the Kglu site is important for elucidating protein molecular function. Due to the time-consuming and expensive limitations of traditional biological experiments, computational-based Kglu site prediction research is gaining more and more attention.

RESULTS

In this paper, we proposed GBDT_KgluSite, a novel Kglu site prediction model based on GBDT and appropriate feature combinations, which achieved satisfactory performance. Specifically, seven features including sequence-based features, physicochemical property-based features, structural-based features, and evolutionary-derived features were used to characterize proteins. NearMiss-3 and Elastic Net were applied to address data imbalance and feature redundancy issues, respectively. The experimental results show that GBDT_KgluSite has good robustness and generalization ability, with accuracy and AUC values of 93.73%, and 98.14% on five-fold cross-validation as well as 90.11%, and 96.75% on the independent test dataset, respectively.

CONCLUSION

GBDT_KgluSite is an effective computational method for identifying Kglu sites in protein sequences. It has good stability and generalization ability and could be useful for the identification of new Kglu sites in the future. The relevant code and dataset are available at https://github.com/flyinsky6/GBDT_KgluSite .

A machine learning and explainable artificial intelligence triage-prediction system for COVID-19

COVID-19 is a respiratory disease caused by the SARS-CoV-2 contagion, severely disrupted the healthcare infrastructure. Various countries have developed COVID-19 vaccines that have effectively prevented the severe symptoms caused by the virus to a certain extent. However, a small section of people continues to perish. Artificial intelligence advances have revolutionized healthcare diagnosis and prognosis infrastructure. In this study, we predict the severity of COVID-19 using heterogenous Machine Learning and Deep Learning algorithms by considering clinical markers, vital signs, and other critical factors. This study extensively reviews various classifier architectures to predict the COVID-19 severity. We built and evaluated multiple pipelines entailing combinations of five state-of-the-art data-balancing techniques (Synthetic Minority Oversampling Technique (SMOTE), Adaptive Synthetic, Borderline SMOTE, SMOTE with Tomek links, and SMOTE with Edited Nearest Neighbor (ENN)) and twelve heterogeneous classifiers such as Logistic Regression, Decision Tree, Random Forest, Support Vector Machine, K-Nearest Neighbors, Naïve Bayes, Xgboost, Extratrees, Adaboost, Light GBM, Catboost, and 1-D Convolution Neural Network. The best-performing pipeline consists of Random Forest trained on Borderline SMOTE balanced data that produced the highest recall of 83%. We deployed Explainable Artificial Intelligence tools such as Shapley Additive Explanations and Local Interpretable Model-agnostic Explanations, ELI5, Qlattice, Anchor, and Feature Importance to demystify complex tree-based ensemble models. These tools provide valuable insights into the significance of critical features in the severity prediction of a COVID-19 patient. It was observed that changes in respiratory rate, blood pressure, lactate, and calcium values were the primary contributors to the increase in severity of a COVID-19 patient. This architecture aims to be an explainable decision-support triaging system for medical professionals in countries lacking advanced medical technology and infrastructure to reduce fatalities.

Combining Resampling Strategies and Ensemble Machine Learning Methods to Enhance Prediction of Neonates with a Low Apgar Score After Induction of Labor in Northern Tanzania

Objective

The goal of this study was to establish the most efficient boosting method in predicting neonatal low Apgar scores following labor induction intervention and to assess whether resampling strategies would improve the predictive performance of the selected boosting algorithms.

Methods

A total of 7716 singleton births delivered from 2000 to 2015 were analyzed. Cesarean deliveries following labor induction, deliveries with abnormal presentation, and deliveries with missing Apgar score or delivery mode information were excluded. We examined the effect of resampling approaches or data preprocessing on predicting low Apgar scores, specifically the synthetic minority oversampling technique (SMOTE), borderline-SMOTE, and the random undersampling (RUS) technique. Sensitivity, specificity, precision, area under receiver operating curve (AUROC), F-score, positive predicted values (PPV), negative predicted values (NPV) and accuracy of the three (3) boosting-based ensemble methods were used to evaluate their discriminative ability. The ensemble learning models tested include adoptive boosting (AdaBoost), gradient boosting (GB) and extreme gradient boosting method (XGBoost).

Results

The prevalence of low (<7) Apgar scores was 9.5% (n = 733). The prediction models performed nearly similar in their baseline mode. Following the application of resampling techniques, borderline-SMOTE significantly improved the predictive performance of all the boosting-based ensemble methods under observation in terms of sensitivity, F1-score, AUROC and PPV.

Conclusion

Policymakers, healthcare informaticians and neonatologists should consider implementing data preprocessing strategies when predicting a neonatal outcome with imbalanced data to enhance efficiency. The process may be more effective when borderline-SMOTE technique is deployed on the selected ensemble classifiers. However, future research may focus on testing additional resampling techniques, performing feature engineering, variable selection and optimizing further the ensemble learning hyperparameters.