Applications of Deep Learning in Bioinformatics and Image Processing

Dear Colleagues,

This Special Issue emphasizes the importance of deep learning in the disciplines of bioinformatics and image processing. Deep learning models are changing how academics approach data-driven research by increasingly being used to analyze and interpret challenging data in a variety of disciplines.

The articles in this Special Issue address a wide range of deep-learning-related topics, including the development of novel algorithms, the application of deep learning to genomics and drug discovery, as well as the classification, segmentation, and feature extraction of images. A number of the studies also look at the integration of deep learning with other technologies such as cloud computing and blockchain.

The articles in this Special Issue show how deep learning is flexible and capable of solving a variety of issues.

Dr. Muhammad Kabir
Guest Editor

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 100 words) can be sent to the Editorial Office for announcement on this website.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Information is an international peer-reviewed open access monthly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 1600 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.

Title: An Enhanced Multi-Localized Attention Feature Extraction Network for Viral Protein Subcellular Localization
Authors: Bakanina Kissanga Grace-Mercure; Yussif Banaamwini Sophyani; Biffon Manyura Momanyi; Lin Ning; Hasan Zulfiqar; Hao Lin
Affiliation: --
Abstract: Accurate prediction of subcellular localization of viral proteins is crucial for understanding their functions and developing effective antiviral drugs. However, this task presents a significant challenge, especially when using expensive and time-consuming classical biological experiments. In this study, we introduced a computational framework, called MuLA, based on deep learning network, which combined multiple local attention modules to enhance feature extraction of protein sequences. The superior performance of the MuLA model has been demonstrated through extensive comparisons with LSTM, CNN, AdaBoost, decision trees, and KNN. It is worth noting that the MuLA could produce the accuracy of 93.66%, specificity of 99.21%, and sensitivity of 89.81%. indicating that MuLA can become an effective tool for predicting virus subcellular localization.

Title: Advancing Early Diagnosis of Cardiac Anomalies through Enhanced ECG Classification: An Experimental Evaluation
Authors: Naba Rahim; Uzair Iqbal; Umar Aftab; Qamar Zaman; Hafiz Tayyab Rauf; Mohamed Sharaf
Affiliation: Centre for Smart Systems, AI and Cybersecurity, Staffordshire University, Stoke-on-Trent ST4 2DE, UK
Abstract: The Electrocardiogram (ECG) is the primary diagnostic tool for determining the numerous cardiac problems that can affect a person. Identification of early diagnosis of cardiovascular issues is very beneficial. Atrial fibrillation (AFIB) is a heart rhythm problem that can cause blood clots in the heart. Myocardial infarction (MI), another name for a heart attack, is a severe medical emergency in which the blood supply to the heart is abruptly interrupted, usually by a blood clot. Early identification of the correlation between cardiac anomalies helps reduce life-threatening hazards, especially in heart stroke variations. The primary motivation for conducting this study is the standardization of retrospective critical analysis, which is crucial for the robust and accurate early diagnosis of several potentially fatal cardiac cases and for safeguarding human life. In this study, ECG classification is done using different machine learning (Support Vector, Naïve Bayes, and Decision Tree) and deep learning methods (Convolutional neural networks or CNN). The study also presents a specific, fine-tuned design for convolutional neural networks (CNN) for verifiable and precise co-relations; MIT-BIH, PTBDB, and SVDB datasets were used for experiments. The results show that Machine learning models did not perform well on arrhythmia classification and provided an accuracy of 71% on the MITDB dataset compared to deep learning models, which performed exceptionally well on both the MITDB dataset and PTBDB dataset with average accuracy and f1-score of 99%. For AFIB and MI correlation, PTBDB, MIT-BIH, and Customized datasets were used with CNN, and all three show exceptional F1-Score of 99% and 97%. In the future, peak detection techniques should be identified and used, which can improve ECG visualization and feature extraction. This study did not consider the segmentation and peak detection, which might have benefited from feature extraction and ECG visualization. Future studies should focus on techniques for peak identification that enhance ECG interpretation. However, there is still potential for enhancements in the proposed approach, such as having the ability to detect brain diseases like stroke or cognitive impairment.

Title: Exploring Human Dietary Sensitivities via BERTopic: A Bioinformatics Approach to Understand Genetic Polymorphisms
Authors: Giovanni Maria De Filippis; Antonio Maria Rinaldi; Cristiano Russo; Cristian Tommasino
Affiliation: Department of Electrical Engineering and Information Technology, University of Naples Federico II, via Claudio 21, 80125, Italy
Abstract: Navigating the vast genomic data on human genetic polymorphisms poses complex analytical challenges in the bioinformatics field. To interpret the health implications due to the interactions of polymorphisms with environmental factors, we need refined, data-driven approaches. Our study applies a topic modeling technique on a comprehensive dataset of abstracts sourced from genomic literature, aiming to comprehend the functional implications of these polymorphisms. The focal areas of our investigation were food tolerances, allergies, diet-induced oxidative stress, and xenobiotics metabolism. By employing BERTopic, a state-of-the-art machine learning model–on PubMed abstracts, we successfully identify the key nutrition-related topics where genetic variation could impact adverse food response and sensitivities towards diet-related oxidative stress. Our research underscores the vital role that machine learning applications, especially deep learning, plays in handling multilayered genomic data, thereby revealing critical associations for personalized dietary therapy and preventive healthcare. The proposed methodology can be replicated across diverse disciplines to effectively decode data complexity.

Title: Improving breast tumor multi-classification from high-resolution histological images with the integration of feature space data augmentation
Authors: Nadia Brancati; Maria Frucci
Affiliation: Institut for High Performance and Networking of the National Research Council of Italy (ICAR-CNR), Naples, Italy
Abstract: To support pathologists in breast tumor diagnosis, deep learning plays a key role in developing histological whole slide image classification methods. However, automatic classification is challenging due to the enormous size of images and the scarcity of representative training data. To tackle these limitations, we propose integrating a deep learning-based breast tumor gigapixel histological image multi-classifier with a high-resolution data augmentation model to process the entire slide by exploring its local and global information and generating its different synthetic versions. The key idea is to perform classification and augmentation in feature latent space, reducing computational cost while preserving the class label of the input. Precisely, we adopt the deep learning-based multi-classification baseline of the BRIGHT Challenge and evaluate the contribution given by a Conditional Generative Adversarial Network-based data augmentation model on the performance of the breast tumor multi-classification for three tumor classes. The proposed method allows achieving an average F1 equal to 69.5, considering only the WSI dataset of the Challenge. The results are comparable to the method presented by the Challenge winner (71.6), who has also been trained on the annotated tumor region dataset of the Challenge.