StackAHTPs: An explainable antihypertensive peptides identifier based on heterogeneous features and stacked learning approach

IET Systems Biology, Journal Year: 2025, Volume and Issue: 19(1)

Published: Jan. 1, 2025

Abstract Hypertension, often known as high blood pressure, is a major concern to millions of individuals globally. Recent studies have demonstrated the significant efficacy naturally derived peptides in reducing pressure. Hypertension one risks associated with cardiovascular disorders and other health problems. Naturally sourced bioactive possessing antihypertensive properties provide considerable potential viable substitutes for conventional pharmaceutical medications. Currently, thorough examination peptide (AHTPs), by using traditional wet‐lab methods highly expensive labours. Therefore, in‐silico approaches especially machine‐learning (ML) algorithms are favourable due saving time cost discovery AHTPs. In this study, novel ML‐based predictor, called StackAHTP was developed predicting accurate AHTPs from sequence only. The proposed method, utilise two types feature descriptors Pseudo‐Amino Acid Composition Dipeptide encode local global hidden information sequences. Furthermore, encoded features serially merged ranked through SHapley Additive explanations (SHAP) algorithm. Then, top fed into three different ensemble classifiers (Bagging, Boosting, Stacking) enhancing prediction performance model. StackAHTPs method achieved superior compare ML (AdaBoost, XGBoost Light Gradient Boosting (LightGBM), Bagging Boosting) on 10‐fold cross validation independent test. experimental outcomes demonstrate that our outperformed existing an accuracy 92.25% F1‐score 89.67% test non‐AHTPs. authors believe research will remarkably contribute large‐scale characterisation accelerate drug process. At https://github.com/ali‐ghulam/StackAHTPs you may find datasets used.

Language: Английский

---

Conotoxins: Classification, Prediction, and Future Directions in Bioinformatics

Toxins, Journal Year: 2025, Volume and Issue: 17(2), P. 78 - 78

Published: Feb. 9, 2025

Conotoxins, a diverse family of disulfide-rich peptides derived from the venom Conus species, have gained prominence in biomedical research due to their highly specific interactions with ion channels, receptors, and neurotransmitter systems. Their pharmacological properties make them valuable molecular tools promising candidates for therapeutic development. However, traditional conotoxin classification functional characterization remain labor-intensive, necessitating increasing adoption computational approaches. In particular, machine learning (ML) techniques facilitated advancements sequence-based classification, prediction, de novo peptide design. This review explores recent progress applying ML deep (DL) research, comparing key databases, feature extraction techniques, models. Additionally, we discuss future directions, emphasizing integration multimodal data refinement predictive frameworks enhance discovery.

Language: Английский

---

Improved in Silico Identification of Protein‐Protein Interactions Using Deep Learning Approach

IET Systems Biology, Journal Year: 2025, Volume and Issue: 19(1)

Published: Jan. 1, 2025

ABSTRACT Protein–protein interactions (PPIs) perform significant functions in many biological activities likewise gene regulation, metabolic pathways and signal transduction. The deregulation of PPIs may cause deadly diseases, such as cancer, autoimmune, pernicious anaemia etc. Detecting can aid elucidating the cellular process's underlying molecular mechanisms contribute to facilitating discovery new proteins for development novel drugs. Although high‐throughput wet‐lab technologies have been matured identify large scale PPI identification; however, traditional experimental methods are costly slow resource intensive. To support techniques, numerous computational approaches emerged identifying solely from protein sequences. However, performance available tools unsatisfactory gaps remain further improvement. In this study, a deep learning‐based model, Deep_PPI, was developed predicting multiple species PPIs. extract features, authors used 21D vector representing 20 kinds' native one special amino acid residue implemented Keras binary profile encoding technique formulate each proteins. use PaddVal strategy equalise length positive negative After extracting fed them into dimension convolutional neural network build final prediction model. proposed Deep_PPI which consider pairs two heads. Finally, concatenated outputs were branches by fully connected layer. efficiency predictor demonstrated both on cross validation tested various datasets, example, that is (Human, C. elegans , E. coli H. sapiens ). model surpassed machine‐learning models existing state‐of‐the‐art methods. will serve valuable tool large‐scale particular provide insights drugs general.

Language: Английский

---

Research on Bitter Peptides in the Field of Bioinformatics: A Comprehensive Review

International Journal of Molecular Sciences, Journal Year: 2024, Volume and Issue: 25(18), P. 9844 - 9844

Published: Sept. 12, 2024

Bitter peptides are small molecular produced by the hydrolysis of proteins under acidic, alkaline, or enzymatic conditions. These can enhance food flavor and offer various health benefits, with attributes such as antihypertensive, antidiabetic, antioxidant, antibacterial, immune-regulating properties. They show significant potential in development functional foods prevention treatment diseases. This review introduces diverse sources bitter discusses mechanisms bitterness generation their physiological functions taste system. Additionally, it emphasizes application bioinformatics peptide research, including establishment improvement databases, use quantitative structure–activity relationship (QSAR) models to predict thresholds, latest advancements classification prediction built using machine learning deep algorithms for identification. Future research directions include enhancing diversifying models, applying generative advance towards deepening discovering more practical applications.

Language: Английский

---

iNP_ESM: Neuropeptide Identification Based on Evolutionary Scale Modeling and Unified Representation Embedding Features

International Journal of Molecular Sciences, Journal Year: 2024, Volume and Issue: 25(13), P. 7049 - 7049

Published: June 27, 2024

Neuropeptides are biomolecules with crucial physiological functions. Accurate identification of neuropeptides is essential for understanding nervous system regulatory mechanisms. However, traditional analysis methods expensive and laborious, the development effective machine learning models continues to be a subject current research. Hence, in this research, we constructed an SVM-based neuropeptide predictor, iNP_ESM, by integrating protein language Evolutionary Scale Modeling (ESM) Unified Representation (UniRep) first time. Our model utilized feature fusion selection strategies improve prediction accuracy during optimization. In addition, validated effectiveness optimization strategy UMAP (Uniform Manifold Approximation Projection) visualization. iNP_ESM outperforms existing on variety evaluation metrics, up 0.937 cross-validation 0.928 independent testing, demonstrating optimal recognition capabilities. We anticipate improved data future, believe that will have broader applications research clinical treatment neurological diseases.

Language: Английский

---

Glypred: Lysine Glycation Site Prediction via CCU–LightGBM–BiLSTM Framework with Multi-Head Attention Mechanism

Journal of Chemical Information and Modeling, Journal Year: 2024, Volume and Issue: 64(16), P. 6699 - 6711

Published: Aug. 9, 2024

Glycation, a type of posttranslational modification, preferentially occurs on lysine and arginine residues, impairing protein functionality altering characteristics. This process is linked to diseases such as Alzheimer's, diabetes, atherosclerosis. Traditional wet lab experiments are time-consuming, whereas machine learning has significantly streamlined the prediction glycation sites. Despite promising results, challenges remain, including data imbalance, feature redundancy, suboptimal classifier performance. research introduces Glypred, site model combining ClusterCentroids Undersampling (CCU), LightGBM, bidirectional long short-term memory network (BiLSTM) methodologies, with an additional multihead attention mechanism integrated into BiLSTM. To achieve this, study undertakes several key steps: selecting diverse types capture comprehensive information, employing cluster-based undersampling strategy balance set, using LightGBM for selection enhance performance, implementing LSTM accurate classification. Together, these approaches ensure that Glypred effectively identifies sites high accuracy robustness. For encoding, five distinct types─AAC, KMER, DR, PWAA, EBGW─were selected broad spectrum sequence biological information. These encoded features were validated information acquisition. address issue highly imbalanced positive negative samples, various algorithms, random undersampling, NearMiss, edited nearest neighbor rule, CCU, evaluated. CCU was ultimately chosen remove redundant nonglycated training data, establishing balanced set enhances model's selection, ensemble algorithm employed reduce dimensionality by identifying most significant features. approach accelerates training, generalization capabilities, ensures good transferability model. Finally, used classifier, structure designed modification from both forward backward directions. prevent overfitting, appropriate regularization parameters dropout rates introduced, achieving efficient Experimental results show achieved optimal provides new insights bioinformatics encourages application similar strategies in other fields. A software tool also developed PyQt5 library, offering researchers auxiliary screening workload improve efficiency. The sets available GitHub: https://github.com/ZBYnb/Glypred.

Language: Английский

---

Alg-MFDL: A multi-feature deep learning framework for allergenic proteins prediction

Analytical Biochemistry, Journal Year: 2024, Volume and Issue: unknown, P. 115701 - 115701

Published: Oct. 1, 2024

Language: Английский

---

Empirical Comparison and Analysis of Artificial Intelligence-Based Methods for Identifying Phosphorylation Sites of SARS-CoV-2 Infection

International Journal of Molecular Sciences, Journal Year: 2024, Volume and Issue: 25(24), P. 13674 - 13674

Published: Dec. 21, 2024

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is a member of the large family with high infectivity and pathogenicity primary pathogen causing global pandemic disease 2019 (COVID-19). Phosphorylation major type protein post-translational modification that plays an essential role in process SARS-CoV-2–host interactions. The precise identification phosphorylation sites host cells infected SARS-CoV-2 will be great importance to investigate potential antiviral responses mechanisms exploit novel targets for therapeutic development. Numerous computational tools have been developed on basis phosphoproteomic data generated by mass spectrometry-based experimental techniques, which can accurately ascertained across whole SARS-CoV-2-infected proteomes. In this work, we comprehensively reviewed several aspects construction strategies availability these predictors, including benchmark dataset preparation, feature extraction refinement methods, machine learning algorithms deep architectures, model evaluation approaches metrics, publicly available web servers packages. We highlighted compared prediction performance each tool independent serine/threonine (S/T) tyrosine (Y) datasets discussed overall limitations current existing predictors. summary, review would provide pertinent insights into exploitation new powerful site tools, facilitate localization more suitable target molecules verification, contribute development therapies.

Language: Английский

---

A Soft Voting Ensemble Model for Hotel Revenue Prediction

International Journal of Economics Finance and Management Sciences, Journal Year: 2024, Volume and Issue: 12(5), P. 258 - 266

Published: Sept. 11, 2024

In recent years, the hotel industry has faced unprecedented opportunities and challenges due to increasing demand for travel business trips. This growth not only presents significant but also brings resource management price setting. Accurate revenue prediction is crucial as it influences pricing strategies allocation. However, traditional models fail capture diversity complexity of data, resulting in inefficient inaccurate predictions. Then, with development ensemble learning, its application emerged an influential research direction. study proposes a soft voting model prediction, which includes six base models: Convolutional Neural Network, K-nearest Neighbors, Linear Regression, Long Short-term Memory, Multi-layer Perceptron, Recurrent Network. Firstly, hyper-parameters are optimized Bayesian optimization. Subsequently, method used aggregate predictions each model. Finally, experimental results on dataset demonstrate that outperforms across key performance metrics, providing managers more accurate tools aid scientific decisions allocation strategies. confirms effectiveness enhancing accuracy forecasts, demonstrating potential strategic planning within modern industry.

Language: Английский

---