International Journal of Bioinformatics Research and Applications (20 papers in press)

Regular Issues

• Meta-Heuristics for Feature Selection: A Comprehensive Survey and Comparative Analysis  
  by Rishika Kumar, Ashish Jain, Inderjeet Kaur 
  Abstract: Feature selection (FS) is a crucial step in pre-processing of data that aims to identify a subset of relevant features from a large pool of available features, while discarding irrelevant or redundant ones. From early 2000s, optimisation heuristic methods have gained popularity as an alternative to traditional FS methods. In the literature, it has been shown that the optimisation heuristics can efficiently search for a subset of relevant features that can represent the data accurately. They are flexible, scalable, and can handle non-differentiable objective functions, making them suitable for FS. In this paper, we comprehensively review those optimisation heuristics that have been developed in last one decade and applied successfully for FS. Each algorithm is elucidated theoretically, providing in-depth explanations of their methodologies. This survey presents difficulties faced by optimisation heuristic FS algorithms and prospective research directions are analysed and highlighted for the benefit of researchers working in this area.
  Keywords: feature selection; optimisation heuristics; data accuracy.
  DOI: 10.1504/IJBRA.2025.10062948
   
  
• Content Based Medical Image Retrieval Using Multi-Feature Extraction and Patch Sorensen Similarity Indexing Technique  
  by K. Saminathan, Amsavalli S, M.Chithra Devi 
  Abstract: In the intricate field of medical imaging, the analysis of image content plays a pivotal role in classification, retrieval, and indexing tasks, as well as in recognising objects and different settings within the image. While innovative, traditional methods typically fail to efficiently and accurately process medical image databases' massive and complicated data. Due to this shortcoming, discrete wavelet coefficients-bag of visual words-contour-local binary pattern (DWC-BoVW-Contour-LBP) relevance fusion was developed. A trimmed mean filter and contrast limited adaptive histogram equalisation (CLAHE) remove noise and boost contrast to optimise the image for feature extraction in this novel method. The system carefully extracts low-level frequency features using discrete wavelet transform (DWT), textural features using local binary pattern (LBP), shape features using contour analysis, and visual features using bag of visual words (BoVW). Pixel image fusion is used to combine various features into a complete picture. Patch Sorensen similarity measurement ranks database photos by query resemblance and selects the top 10 most similar images. The algorithm's precision, F-score, and recall were superior in the TCIA-CT database, showing a substantial progress in content-based medical image retrieval (CBMIR).
  Keywords: bag of visual words; BoVW; discrete wavelet transform; DWT; image retrieval.Sorensen; similarity indexing technique.
  DOI: 10.1504/IJBRA.2025.10063649
   
  
• Computational promoter analysis of ovine lipogenic genes reveals insulin as a major factor potentially responsible for their synergistic transcription  
  by Kristian Christos Kailis, Iosif Bizelis, Panagiota Koutsouli, George Laliotis 
  Abstract: De novo lipogenesis (DNL) is an important metabolic pathway in ruminants and non-ruminants. In sheep, cytosolic dehydrogenases of glucose-6-phosphate (G6PD), 6-phosphorogluconic (6PGD), isocitric (IDH1) and malic acid (ME1) contribute to the required NADPH for DNL. A synergic action of these enzymes has been reported, but the common regulatory mechanism is unknown. The study aimed to identify: 1) the promoter region of these genes; 2) any potential regions for transcription factor binding; 3) potential common transcription factors; 4) differences between ruminant and non-ruminant counterparts. Results showed structural differences in the promoter region among species. Ten transcription factors were found to be commonly present, including USF1 and HNF-3?, which are involved in lipogenesis and are considered as insulin-dependent. Therefore, the insulin-USF-HNF triptych may have an impact on the synergistic action of the studied genes since both USF and HNF expression are insulin-dependent. Accordingly, any stimulus that alters insulin could alter the expression of the studied genes through USF and HNF accumulation. Overall, this study provides valuable insights into the transcription process of these genes and the lipogenesis pathway in sheep.
  Keywords: lipogenesis; fatty acids; ruminants; sheep; metabolism; NADPH; genomics; transcription factors.
  DOI: 10.1504/IJBRA.2025.10063779
   
  
• The Estimation of Statistical Features from VMD Levels for Automated Sleep Apnoea Classification  
  by Suchetha M, Smruthy A, EDWIN DHAS D, Sehastrajit S, Ziani Said 
  Abstract: The sleep-related diseases are common nowadays. It is affecting around 30% of the total population all over the world. The main reasons for sleep apnoea syndrome are the lack of exercise and obesity. It is important to screen the sleep apnoea because it indirectly affects the cardiovascular and intelligence quotient (IQ) functions. In this proposed work, we are introducing a novel classification of apnoea and healthy subjects by using the Variational Mode Decomposition (VMD) algorithm. The main intention of this work is to extract the different statistical features from the decomposed Electrocardiogram (ECG) modes and classify the features that are extracted from the decomposed modes using the support vector machine (SVM) classifier model. Our proposed work attained an accuracy of 97.56% in the classification of sleep apnoea.
  Keywords: Variational Mode Decomposition; Sleep Apnea; Electro Cardiogram; Support Vector Machine; Intelligent Quotient.
  DOI: 10.1504/IJBRA.2025.10063791
   
  
• Identifying Molecular Subtypes of Breast Cancer using Single Cell RNA-Seq Data Integration and Random Forest Classification  
  by Peter Jerome Ishmael Paulino, Muhammad Sufyan 
  Abstract: Single-cell RNA sequencing (scRNA-seq) has been invaluable in advancing our understanding of various cancers, including breast cancer. The extensive analysed scRNA-seq data from multiple independent breast cancer studies to build an integrated single cell gene expression atlas encompassing over 60,000 cells. Unsupervised clustering and classification algorithms including t-SNE, UMAP, and random forest were applied to identify molecular subtypes and classify new tumour samples. Integrated analysis identified six major breast cancer subtypes consistent with known luminal, HER2-enriched, and basal-like classifications. Random forest classification using a panel of discriminative genes achieved over 90% accuracy in classifying held-out tumour samples into known subtypes. Further substructure within subtypes revealed novel candidate cell states. The study also demonstrated the feasibility and advantages of integrating multiple scRNA-seq datasets to generate a comprehensive breast cancer atlas. The results of this study provide insights into breast cancer biology with potential applications in precision oncology.
  Keywords: breast cancer; single-cell RNA-seq; scRNA-seq; data integration; molecular subtypes; random forest; tumour heterogeneity.
  DOI: 10.1504/IJBRA.2024.10064030
   
  
• IRNN-SS: Deep Learning for Optimized Protein Secondary Structure Prediction through PROMOTIF and DSSP Annotation Fusion  
  by Mukhtar Sofi, M. Arif Wani 
  Abstract: DSSP stands as a foundational tool in the domain of protein secondary structure prediction, yet it encounters notable challenges in accurately annotating irregular structures, such as -turns and -turns, which constitute approximately 25-30% and 10-15% of protein turns, respectively. This limitation arises from DSSP's reliance on hydrogen-bond analysis, resulting in annotation gaps and reduced consensus on irregular structures. Alternatively, PROMOTIF excels at identifying these irregular structure annotations using phi-psi information. Despite their complementary strengths, previous methodologies utilised DSSP and PROMOTIF separately, leading to disparate prediction methods for protein secondary structures, hampering comprehensive structure analysis crucial for drug development. In this work, we bridge this gap using an annotation fusion approach, combining DSSP structures with beta, and gamma turns. We introduce IRNN-SS, a model employing deep inception and bidirectional gated recurrent neural networks, achieving 77.4% prediction accuracy on benchmark datasets, outpacing current models.
  Keywords: protein secondary structure; PSS; beta-turns; gamma-turns; prediction; deep learning.
  DOI: 10.1504/IJBRA.2025.10064216
   
  
• Feature Analytics of Asthma Severity Levels for Bioinformatics Improvement Using Gini Importance  
  by Temitope Elizabeth Ogunbiyi, Micheal A. Adegoke, Oluwatobi A. Abe, Joseph A. Ojo 
  Abstract: In the context of asthma severity prediction, this study delves into the feature importance of various symptoms and demographic attributes. Leveraging a comprehensive dataset encompassing symptom occurrences across varying severity levels, this investigation employs visualisation techniques, such as stacked bar plots, to illustrate the distribution of symptomatology within different severity categories. Additionally, correlation coefficient analysis is applied to quantify the relationships between individual attributes and severity levels. Moreover, the study harnesses the power of Random Forest and the Gini importance methodology, essential tools in feature importance analytics, to discern the most influential predictors in asthma severity prediction. The experimental results bring to light compelling associations between certain symptoms, notably runny-nose and nasal-congestion, and specific severity levels, elucidating their potential significance as pivotal predictive indicators. Conversely, demographic factors, encompassing age groups and gender, exhibit comparatively weaker correlations with symptomatology. These findings underscore the pivotal role of individual symptoms in characterising asthma severity, reinforcing the potential for feature importance analysis to enhance predictive models in the realm of asthma management and bioinformatics.
  Keywords: bioinformatics; asthma; severity prediction; feature importance; machine learning.
  DOI: 10.1504/IJBRA.2025.10064255
   
  
• Alzheimer's Disease Classification using Hybrid Alex-ResNet-50 Model  
  by E. Semmalar, R. Shoba Rani 
  Abstract: Alzheimer’s disease (AD), a leading cause of dementia and mortality, presents a growing concern due to its irreversible progression and the rising costs of care. Early detection is crucial for managing AD, which begins with memory deterioration caused by the damage to neurons involved in cognitive functions. Although incurable, treatments can manage its symptoms. This study introduces a hybrid AlexNet+ResNet-50 model for AD diagnosis, utilising a pre-trained convolutional neural network (CNN) through transfer learning to analyse MRI scans. This method classifies MRI images into Alzheimer's disease (AD), moderate cognitive impairment (MCI), and normal control (NC), enhancing model efficiency without starting from scratch. Incorporating transfer learning allows for refining the CNN to categorise these conditions accurately. Our previous work also explored atlas-based segmentation combined with a U-Net model for segmentation, further supporting our findings. The hybrid model demonstrates superior performance, achieving 94.21% accuracy in identifying AD cases, indicating its potential as a highly effective tool for early AD diagnosis and contributing to efforts in managing the disease's impact.
  Keywords: Alzheimer’s disease; AD; segmentation; classification; CNN; AlexNet; U-Net; ResNet-50; moderate cognitive impairment.
  DOI: 10.1504/IJBRA.2025.10064352
   
  
• Optimisation with Deep Learning for Leukaemia Classification in Federated Learning  
  by Smritilekha Das, PADMANABAN K 
  Abstract: The most common kind of blood cancer in people of all ages is leukaemia. The fractional mayfly optimisation (FMO) based DenseNet is proposed for the identification and classification of leukaemia in federated learning (FL). Initially, the input image is pre-processed by adaptive median filter (AMF). Then, cell segmentation is done using the Scribble2label. Afterthat, image augmentation is accomplished. Finally, leukaemia classification is accomplished utilising DenseNet, which is trained using the FMO. Here, the FMO is devised by merging the mayfly algorithm (MA) and the fractional concept (FC). Following local training, the server performs local updating and aggregation using a weighted average by RV coefficient. The results showed that FMO-DenseNet attained maximum accuracy, true negative rate (TNR) and true positive rate (TPR) of 94.3%, 96.5% and 95.3%. Moreover, FMO-DenseNet gained minimum mean squared error (MSE) and root mean squared error (RMSE) of 5.7%, 9.2% and 30.4%.
  Keywords: leukaemia; federated learning; fractional concept; mayfly algorithm; DenseNet.
  DOI: 10.1504/IJBRA.2024.10064487
   
  
• Deep Learning Approach using Modified DarkNet-53 for Renal Cell Carcinoma Grading  
  by G. Sathish Kumar, G. Uma Maheshwari, C. Selvan, M. Nagasuresh, Rasi D, Swathypriyadharsini Palaniswamy, Sathish Kumar Danasegaran 
  Abstract: An accurate and effective diagnostic procedures are required for appropriate treatment planning for renal cell carcinoma, the most frequent form of kidney cancer. Using fusion module a network dubbed Modified Darknet (MDNet) was developed for image-based small-target detection. We built MDNet on top of a modified version of Darknet53, which itself a scale matching approach, to increase its speed and accuracy. By combining the results of several convolutional neural network (CNN) models, the ensemble structure improves classification accuracy. The effectiveness of a classification algorithm using kidney histopathology pictures dataset is measured in accuracy, precision, recall, sensitivity, specificity and f1-score. The results show that the ensemble deep learning method outperforms both standalone CNN models and more conventional machine learning techniques in RCC classification. Overall grade classification accuracy of 98.9%, a sensitivity of 98.2%, and a high classification specificity of 98.7%, in distinguishing tissues.
  Keywords: Modified Darknet; Convolutional Neural Network; Ensemble Deep Learning; Kidney Cancer; Renal Cell Carcinoma; Whole Slide Images.
  DOI: 10.1504/IJBRA.2025.10064488
   
  
• Automatic Pectoral Muscles and Artifacts Removal in Mammogram Images for Improved Breast Cancer Diagnosis  
  by Saguna Ingle, Amarsinh Vidhate, Sangita Chaudhari 
  Abstract: Breast cancer is leading cause of mortality among women compared to other types of cancers. Hence, early breast cancer diagnosis is crucial to the success of treatment. Various pathological and imaging tests are available for the diagnosis of breast cancer. However, it may introduce errors during detection and interpretation, leading to false-negative and false-positive results due to lack of pre-processing of it. To overcome this issue, we proposed a effective image pre-processing technique-based on Otsu's thresholding and Single-Seeded Region Growing (SSRG) to remove artifacts and segment the pectoral muscle from breast mammograms. To validate the proposed method, a publicly available MIAS dataset has utilised. The experimental finding showed that proposed technique improved 18% breast cancer detection accuracy compared to existing methods. The proposed methodology works efficiently for artifact removal and pectoral segmentation at different shapes and nonlinear patterns.
  Keywords: breast cancer; artifacts; pectoral muscle; image processing; mammogram; image enhancement.
  DOI: 10.1504/IJBRA.2024.10064495
   
  
• Autism Spectrum Disorder Detection using Machine Learning Techniques  
  by Abdelhakim Ridouh, Fayçal IMEDJDOUBEN, Sarra Mahi 
  Abstract: Autism is a developmental disorder that occurs in early childhood and affects communication and social interaction; it includes specific and recurring patterns of behaviour. There is no specific cause for autism and there is no direct treatment for it. Symptoms appear in early childhood and early diagnosis allows for a rise in the recovery rate. In this paper, we present a method to characterise, identify, and classify some ASD data by using machine learning methods such as SVM, DT, NB, and KNN. The studies are carried out on some real ASD data collected from an international database of three categories (children, adolescents, and adults). To enrich the database, we collected more samples from Algeria. The estimation of the best value of parameters for each distribution is achieved by calculating three main parameters illustrated by the confusion matrix. The results illustrate the effectiveness of the proposed method with the best precision.
  Keywords: autism detection; machine learning techniques.
  DOI: 10.1504/IJBRA.2024.10064736
   
  
• Distance-based Contact Maps Prediction for RNA Bases using Deep Neural Networks and Single Sequence Features  
  by Mahmood Rashid, Kuldip Paliwal 
  Abstract: RNA molecules play critical roles in various biological processes which are predominantly governed by their secondary and tertiary structures. The secondary structure of RNA help us understand the functional behaviours and regulatory mechanisms of the RNA molecules. Although the experimental methods can determine highly accurate structures, those methods are expensive, time consuming and labour intensive. As a result, the gap between the number of known sequences and the number of known structures are increasing rapidly. The recent advancements in artificial intelligence and increasing number of known structures encourage researchers build deep learning models to predict RNA structures aiming to reduce this gap. Towards finding an efficient deep learning architecture, we implemented VGG16, VGG19, AlexNet, ResNet and GoogLeNet architecture based convolutional neural networks and trained them on single sequence RNA features. Along with the superior performance over other architectures, we found that the GoogLeNet based model improves the F1 scores (validation F1 = 0.74 and test F1 = 0.66) in comparison to the state-of-the-art F1 scores (validation F1 = 0.71 and test F1 = 0.64) for both validation and test datasets.
  Keywords: RNA structure prediction; RNA contact maps; arti?cial neural networks; deep learning architectures; GoogLeNet; ResNet and VGG.
  DOI: 10.1504/IJBRA.2024.10064821
   
  
• HDAC Inhibitors and their Potential towards Cancer Treatment  
  by Sanjay Kumar Choubey, Sachin Kumar, Medha Kumari 
  Abstract: Histone deacetylases (HDACs) play a key role in chromatin structure modulation through deacetylation of histones leading to formation of highly compact DNA-histone complex. HDACs have been reported to be implicated in multiple types of cancers. Blocking the activities of histone deacetylases will help to overcome gene repression pressure and it would be possible to check the incessant growth of cells in tumour. Therefore the interest has been developed to design the HDAC inhibitors and their analogues and histone deacetylases are now considered as potential targets for their wide distribution in various forms of cancer. HDAC inhibitors display their role by regulating cyclin dependent kinases (cdK), inducing p21 and various preapoptotic genes like Bax, Bak, repressing the activities of growth factors like VEGF, repressing transcription factor HIF-1 facilitating arrest of cell cycle, modulating various signalling pathways like STAT signalling, AMPK signalling, inducing cell adhesion molecule E-cadherin.
  Keywords: cyclin dependent kinase; histone deacetylase; carcinogenesis; HDAC inhibitor.
  DOI: 10.1504/IJBRA.2025.10065088
   
  
• Computational Analysis of Alkoxy-Azoxybenzene Liquid Crystals: A Comparative Investigation with Experimental Data for Bioinformatics Applications  
  by Sushma M, Mahadev J, Manju V. V, Nandaprakash M. B, Somashekar R 
  Abstract: Through computational modelling, we have gained valuable insights into the homologous series of liquid crystalline materials. Our study involved comparing the computational results with reported experimental values for several members of the series. We focused on various parameters, including lattice energy, orientational order parameter, moduli, stress-strain behaviour, Helmholtz free energy, orientational distribution function, zero-point energy, and molecular polarisabilities. The primary motivation behind this study was to unravel the intricate inter- and intra-molecular interactions that govern the range and nature of mesophases observed in these compounds. We are excited to report that our results align with this objective, highlighting the significance of our findings in this direction. Knowledge of these compounds finds applications in sensitive nucleic acid detection, label-free protein analysis, and the development of biocompatible sensors for real-time cellular monitoring.
  Keywords: liquid crystal; odd-even effect; elastic moduli.
  DOI: 10.1504/IJBRA.2025.10065091
   
  
• A Novel Linear Discriminant Analysis Based Classification of R-peaks in Different ECG Signal Datasets  
  by Varun Gupta 
  Abstract: In the current scenario, there is a need to develop efficient pre-processing and classification techniques which can form the basis of an automated health monitoring system. In this paper, independent component analysis (ICA) is proposed to be used for electrocardiogram (ECG) signal processing as reported by the same authors, who found it to yield better results that time for limited datasets. Here, it has been applied on a variety of datasets, viz., real and standard and the obtained results are compared with those obtained using another widely used and reported technique, viz., adaptive notch filter (ANF) in the literature. For classification, linear discriminant analysis (LDA) is proposed to be used as it performs multi-class classification tasks better. The obtained results demonstrate the utility of the proposed methodology for bioinformatics community, especially during critical heart surgeries and designing of evolving healthcare systems in future.
  Keywords: electrocardiogram; ECG; adaptive notch filter; ANF; independent component analysis; ICA; linear discriminant analysis; LDA; signal-to-noise ratio; SNR.
  DOI: 10.1504/IJBRA.2025.10065193
   
  
• A Novel Approach for Early Detection and Grading of Diabetic Retinopathy by using Ensemble Model  
  by Riddhi Parasnaik, Anvita Agarkar, Raashi Jatakia, Gajanan Nagare 
  Abstract: This study investigates the factors driving HR professionals' intention to adopt AI in talent acquisition in the Indian IT industry by adopting a mixed technology-organization-environment (TOE) and task-technology fit (TTF) model. We administered a survey instrument on 459 HR professionals including talent acquisition executives randomly selected from various IT firms located in major Indian cities. The PLS-SEM results revealed that the perception of cost effectiveness, relative advantage, HR readiness, top management support and competitive pressure significantly influence the adoption intention of HR professionals of the Indian IT companies. The findings of the study would help understand the factors that influence HR managers' decisions to adopt AI in talent acquisition process. Further, the study contributes to the existing adoption theories by integrating TOE and TTF models to HR contexts and offers actionable insights for practicing managers of the organisations aiming to adopt AI in the recruitment process.
  Keywords: digital transformation; artificial intelligence; AI; talent acquisition; technology-organisation-environment model; task technology fit model; India.
  DOI: 10.1504/IJBRA.2025.10065195
   
  
• Characterising the Cardioprotective Potential of Sida Rhombifolia, Polygonum Chinense and Phyla Nodiflora Aqueous Extracts: Investigating its Effect on Foam Cell Formation  
  by Xiao Wei Lee, Wei Sheng Siew, Siau Hui Mah, Wei Hsum Yap 
  Abstract: Cardiovascular diseases represent one of the leading causes of mortality. Studies have shown that medicinal plants with anti-inflammatory and antioxidant activities are potential cardioprotective agents. This study aimed to determine cardioprotective potential of Sida rhombifolia, Polygonum chinense and Phyla nodiflora in inhibiting macrophage foam cells formation and its regulatory mechanisms. The findings showed that S. rhombifolia and P. nodiflora have minimal cytotoxicity effect on THP-1 macrophages, however P. chinense exhibited cytotoxic effect with an IC50 of 11.83
  Keywords: atherosclerosis; foam cell; network pharmacology.
  DOI: 10.1504/IJBRA.2025.10065324
   
  
• A Comparative Study on the Classification of SARS-CoV-2 Variants from Biosequence Images using Pre-Trained Deep Learning Models  
  by Shahina K, Biji C. L, Achuthsankar S. Nair 
  Abstract: Coronavirus disease has raised serious health concern across the globe. Identification of severe acute respiratory syndrome corona virus 2 (SARS-CoV-2) variants are indeed a concern in controlling its spread. SARS-CoV-2 variants are classified based on the variation in its genomic sequences. Alpha, beta, delta, gamma and omicron were reported as the most deleterious variants. Genome sequence can be represented uniquely using chaos game representation (CGR) images. A large-scale genome sequence dataset, belonging to the five categories of these variant were retrieved from GISAID. An attempt was made to compile benchmark CGR images of 25,000 SARS-CoV-2 variants genomic sequences. The present study aims to compare the performance of different pre-trained deep learning models in classifying SARS-CoV-2 variants from its CGR images. VGG16, VGG19, ResNet50, InceptionV3, Xception, InceptionResNetV2 and MobileNetV2 were the models used for the study. SARS-CoV-2 variant detection was found effective with VGG19 with an accuracy of 94%. Data augmentation techniques were also applied on the CGR images of biosequences and it was found that data augmentation methods decreased the accuracy of different transfer learning models.
  Keywords: genome sequence; deep learning; SARS-CoV-2 variants; chaos game representation; transfer learning; classification; COVID-19.
  DOI: 10.1504/IJBRA.2025.10065325
   
  
• Skin Cancer Classification using Ensemble Classification Model with Improved Deep Joint Segmentation  
  by Jinu P. Sainudeen, Sathyalakshmi S 
  Abstract: We present a six-phase skin cancer classification model based on Improved Deep Joint Segmentation (IDJS) in this work. The pre-processed image is segmented using IDJS in the second phase, after contrast enhancement with assistance from Contrast Limited Adaptive Histogram Equalization (CLAHE) in the first phase. The features of GLCM, CCF, LGIP, and Median Ternary Pattern (MTP) are retrieved in the third phase. Data augmentation for the extracted features is carried out in the fourth phase. The fifth phase is ensemble classification using the Deep Maxout, LSTM, and CNN based on the enhanced data. To determine the final classified label, the enhanced score level fusion receives the output scores from these classifiers. While the RF is 0.9171, Deep Maxout is 0.9382, LSTM is 0.9362, Bi-GRU is 0.8150, RNN is 0.8687, CNN is 0.9382, TL-GOOGLENET is 0.9134, and KNN is 0.9328, respectively, the accuracy of the Ensemble approach is 0.9689.
  Keywords: DL; Skin cancer; segmentation; Classification; Recommendation.
  DOI: 10.1504/IJBRA.2025.10065333