Anjan Chowdhury

The outbreak of coronavirus disease 2019 (COVID-19), caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has emerged as one of the most significant global health crises in recent history. Coronaviruses are a diverse group of RNA viruses classified into alpha, beta, gamma, and delta genera, with SARS-CoV-2 belonging to the beta-coronavirus family. The virus exhibits high transmissibility and causes a wide spectrum of clinical manifestations ranging from mild respiratory symptoms to severe complications such as acute respiratory distress syndrome, multi-organ failure, and death, particularly among elderly and immunocompromised individuals. Structurally, SARS-CoV-2 possesses a large single-stranded RNA genome encoding major structural proteins, including spike (S), envelope (E), membrane (M), and nucleocapsid (N) proteins, which play critical roles in host-cell recognition and viral infection. Understanding the molecular mechanisms of virus–host interactions, especially protein–protein interactions (PPIs), is essential for uncovering viral pathogenesis and identifying potential therapeutic targets. Traditional experimental techniques for PPI detection, such as yeast two-hybrid and affinity purification methods, are often expensive, labor-intensive, and prone to inaccuracies. Consequently, computational approaches based on machine learning (ML) and deep learning (DL) have gained significant attention for efficient and scalable PPI prediction. These methods use diverse biological information, including protein sequences, structural features, genomic data, Gene Ontology annotations, and interaction networks, to model complex biological relationships. This survey reviews computational approaches to PPI prediction, highlighting ML- and DL-based techniques, methodological advances, performance evaluation practices, and limitations that affect benchmark comparability. It also discusses biological databases and data sources commonly used in PPI studies and explicitly considers how models trained in coronavirus-centered settings may generalize to other viral families with different mechanisms of host interaction.

Background Intracranial steno-occlusive lesions are characteristic of moyamoya disease (MMD), but increasing reports of extracranial vascular involvement suggest a possible systemic arteriopathy. Methods We conducted a PRISMA-guided systematic review (1974–May 2025) to identify extracranial vascular involvement in classical MMD. Seventy-four studies were included. Because the search was limited to English-language publications, language bias may be present. A subset of 85 patients from 46 studies was used for exploratory multilabel machine-learning and deep-learning modeling based on clinical variables. Results Seventy-four studies comprising 143 patients (59.4% female) were included; mean age was 24.9 years. Bilateral intracranial disease predominated. Among patients with isolated extracranial involvement, coronary lesions were most frequent (33.1%), followed by renal (31.5%), pulmonary (14.6%), external carotid (10.8%), vertebral (4.6%), and celiac/mesenteric (3.1%) involvement; other rare single-vessel lesions accounted for 2.3%. Multisite involvement occurred in 9.1%. Of 25 genotyped patients, 23 (92.0%) carried RNF213 variants. Mortality was highest in the pulmonary subgroup (31.6%). Logistic regression and random forest showed the best overall balance of accuracy, F1 score, and ROC area, whereas one-dimensional convolutional neural networks were the strongest deep-learning models, although deep learning overall underperformed machine learning. Feature selection identified vascular risk factors and bilateral MMD as the strongest predictors. Conclusions Extracranial vascular involvement in MMD appears age- and vascular bed-specific. These findings are consistent with, but do not establish, a systemic arteriopathy framework and support further prospective validation.

The outbreak of coronavirus disease 2019 (COVID-19), caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has emerged as one of the most significant global health crises in recent history. Coronaviruses are a diverse group of RNA viruses classified into alpha, beta, gamma, and delta genera, with SARS-CoV-2 belonging to the beta-coronavirus family. The virus exhibits high transmissibility and causes a wide spectrum of clinical manifestations ranging from mild respiratory symptoms to severe complications such as acute respiratory distress syndrome, multi-organ failure, and death, particularly among elderly and immunocompromised individuals. Structurally, SARS-CoV-2 possesses a large single-stranded RNA genome encoding major structural proteins, including spike (S), envelope (E), membrane (M), and nucleocapsid (N) proteins, which play critical roles in host cell recognition and viral infection. Understanding the molecular mechanisms of virus–host interactions, especially protein–protein interactions (PPIs), is essential for uncovering viral pathogenesis and identifying potential therapeutic targets. Traditional experimental techniques for PPI detection, such as yeast two-hybrid and affinity purification methods, are often expensive, labor-intensive, and prone to inaccuracies. Consequently, computational approaches based on machine learning and deep learning have gained significant attention for efficient and scalable PPI prediction. These methods utilize diverse biological information, including protein sequences, structural features, genomic data, gene ontology annotations, and interaction networks, to model complex biological relationships. This survey provides a comprehensive review of computational approaches for PPI prediction, highlighting both machine learning- and deep learning-based techniques, along with their methodological advancements and performance evaluations. Furthermore, the survey discusses major biological databases and data sources commonly employed in PPI studies, offering insights into current challenges and future directions in computational PPI prediction research.

Mitochondrial disorders are a diverse group of inherited diseases caused by defects in oxidative phosphorylation, the primary way cells produce energy. These disorders often affect multiple organ systems, particularly those with high energy demands, such as the brain, muscles, and heart. Although increasingly recognized worldwide, mitochondrial diseases are still underdiagnosed in resource-limited areas. This prospective observational study aimed to describe the clinical, biochemical, imaging, electrophysiological, and genetic features of 48 patients suspected of having mitochondrial disease at a tertiary care center in Eastern India. Participants were classified into four main clinical syndromes: mitochondrial encephalomyopathy with lactic acidosis and stroke-like episodes; Leber hereditary optic neuropathy; Leigh syndrome; and chronic progressive external ophthalmoplegia. The average age was 20.4 years (standard deviation 12.3; range 5–60), with a slight male predominance. Molecular testing confirmed known pathogenic mutations in mitochondrial transfer RNA genes, mitochondrial respiratory complex genes, and ATP synthase subunits, depending on the syndrome. Patients exhibited characteristic abnormalities on brain imaging, including stroke-like lesions in the temporal and parietal lobes or symmetrical involvement of the basal ganglia. Biochemical tests revealed elevated lactate levels in blood and cerebrospinal fluid. Electroencephalography often showed slowing or epileptiform discharges in those with brain involvement. Machine learning techniques helped identify clinical patterns, demonstrating their potential to aid in diagnosis. We offer one of the most comprehensive overviews of mitochondrial diseases in South Asia to date, highlighting the need for accessible molecular diagnostics and the usefulness of computational tools in understanding complex disease phenotypes.

Graph Neural Networks (GNNs) are a popular machine learning framework for solving various graph processing applications. This framework exploits both the graph topology and the feature vectors of the nodes. One of the important applications of GNN is in the semi-supervised node classification task. The accuracy of the node classification using GNN depends on (i) the number and (ii) the choice of the training nodes. In this article, we demonstrate that increasing the training nodes by selecting nodes from the same class that are spread out across non-contiguous subgraphs, can significantly improve the accuracy. We accomplish this by presenting a novel input intervention technique that can be used in conjunction with different GNN classification methods to increase the non-contiguous training nodes and, thereby, improve the accuracy. We also present an output intervention technique to identify misclassified nodes and relabel them with their potentially correct labels. We demonstrate on real-world networks that our proposed methods, both individually and collectively, significantly improve the accuracy in comparison to the baseline GNN algorithms. Both our methods are agnostic. Apart from the initial set of training nodes generated by the baseline GNN methods, our techniques do not need any other extra knowledge about the classes of the nodes. Thus, our methods are modular and can be used as pre-and post-processing steps with many of the currently available GNN methods to improve their accuracy.