Titre :	Machine learning for the prediction of biological sequence structures
Type de document :	document multimédia
Auteurs :	Rached Yagoubi, Auteur ; Aabdelouahab Moussaoui, Directeur de thèse
Editeur :	Laghouat : Université Amar Telidji - Département de mathématiques
Année de publication :	2025
Importance :	86 p.
Accompagnement :	1 disque optique numérique (CD-ROM)
Langues :	Anglais
Catégories :	THESES :10 informatique
Mots-clés :	Bioinformatics  Protein structural classes prediction  Machine learning  Deep learning  Deep neural network (DNN)  Convolutional neural network (CNN)
Résumé :	Accurately predicting protein structural classes plays a crucial role in bioinformatics, facilitating insights into protein function and interactions. However, predicting these classes remains a challenge, particularly when working with proteins that exhibit low sequence similarity. This thesis aims to address these challenges by exploring and comparing machine learning and deep learning approaches for protein structural class prediction. The first contribution involves the use of Recursive Feature Elimination (RFE) to select key features from protein sequences, followed by the application of a Support Vector Machine (SVM) classifier, resulting in improved accuracy by focusing on the most relevant features. The second contribution introduces a Feedforward Deep Neural Network (DNN), which further enhances prediction accuracy by learning complex representations from the selected features. The third and most successful approach employs a Convolutional Neural Network (CNN), which automatically extracts and classifies protein structural classes from the predicted secondary structure sequences. The results demonstrate that the CNN-based approach achieves the highest overall accuracy across four low similarity benchmark datasets, including 25PDB, 640, 1189, and FC699, making it a highly effective method for protein structural class prediction. This work highlights the potential of deep learning models in advancing the field and provides a foundation for future research in protein structural classes prediction.
note de thèses :	Thèse de doctorat en informatique

Machine learning for the prediction of biological sequence structures [document multimédia] / Rached Yagoubi, Auteur ; Aabdelouahab Moussaoui, Directeur de thèse . - Laghouat : Université Amar Telidji - Département de mathématiques, 2025 . - 86 p. + 1 disque optique numérique (CD-ROM).
Langues : Anglais

Catégories :	THESES :10 informatique
Mots-clés :	Bioinformatics  Protein structural classes prediction  Machine learning  Deep learning  Deep neural network (DNN)  Convolutional neural network (CNN)
Résumé :	Accurately predicting protein structural classes plays a crucial role in bioinformatics, facilitating insights into protein function and interactions. However, predicting these classes remains a challenge, particularly when working with proteins that exhibit low sequence similarity. This thesis aims to address these challenges by exploring and comparing machine learning and deep learning approaches for protein structural class prediction. The first contribution involves the use of Recursive Feature Elimination (RFE) to select key features from protein sequences, followed by the application of a Support Vector Machine (SVM) classifier, resulting in improved accuracy by focusing on the most relevant features. The second contribution introduces a Feedforward Deep Neural Network (DNN), which further enhances prediction accuracy by learning complex representations from the selected features. The third and most successful approach employs a Convolutional Neural Network (CNN), which automatically extracts and classifies protein structural classes from the predicted secondary structure sequences. The results demonstrate that the CNN-based approach achieves the highest overall accuracy across four low similarity benchmark datasets, including 25PDB, 640, 1189, and FC699, making it a highly effective method for protein structural class prediction. This work highlights the potential of deep learning models in advancing the field and provides a foundation for future research in protein structural classes prediction.
note de thèses :	Thèse de doctorat en informatique