Using parenclitic networks on phaeochromocytoma and paraganglioma tumours provides novel insights on global DNA methylation DOI Creative Commons

Dimitria Brempou,

Bertille Montibus, Louise Izatt

et al.

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: Dec. 2, 2024

Abstract Despite the prevalence of sequencing data in biomedical research, methylome remains underrepresented. Given importance DNA methylation gene regulation and disease, it is crucial to address need for reliable differential methods. This work presents a novel, transferable approach extracting information from data. Our agnostic, graph-based pipeline overcomes limitations commonly used techniques addresses “small n, big k” problem. Pheochromocytoma Paraganglioma (PPGL) tumours with known genetic aetiologies experience extreme hypermethylation genome wide. To highlight effectiveness our method candidate discovery, we present first phenotypic classifier PPGLs based on achieving 0.7 ROC-AUC. Each sample represented by an optimised parenclitic network, graph representing deviation sample’s expected non-aggressive patterns. By meaningful topological features, dimensionality and, hence, risk overfitting reduced, samples can be classified effectively. using explainable classification method, this case logistic regression, key CG loci influencing decision identified. provides insights into molecular signature aggressive propose candidates further research. network implementation improves potential utility offers effective complete studying such datasets.

Language: Английский

Comparative Analysis of Multi-Omics Integration Using Graph Neural Networks for Cancer Classification DOI Creative Commons

Fadi Alharbi,

Aleksandar Vakanski, Boyu Zhang

et al.

IEEE Access, Journal Year: 2025, Volume and Issue: 13, P. 37724 - 37736

Published: Jan. 1, 2025

Recent studies on integrating multiple omics data highlighted the potential to advance our understanding of cancer disease process. Computational models based graph neural networks and attention-based architectures have demonstrated promising results for classification due their ability model complex relationships among biological entities. However, challenges related addressing high dimensionality complexity in multi-omics data, as well constructing structures that effectively capture interactions between nodes, remain active areas research. This study evaluates network (MO) integration graph-convolutional (GCN), graph-attention (GAT), graph-transformer (GTN). Differential gene expression LASSO (Least Absolute Shrinkage Selection Operator) regression are employed reducing feature selection; hence, developed referred LASSO-MOGCN, LASSO-MOGAT, LASSO-MOGTN. Graph constructed using sample correlation matrices protein-protein interaction investigated. Experimental validation is performed with a dataset 8,464 samples from 31 types normal tissue, comprising messenger-RNA, micro-RNA, DNA methylation data. The show outperformed trained single where LASSO-MOGAT achieved best overall performance, an accuracy 95.9%. findings also suggest correlation-based enhance models' identify shared cancer-specific signatures across patients comparison networks-based structures. code used this available link (https://github.com/FadiAlharbi2024/Graph_Based_Architecture.git).

Language: Английский

Citations

2
Semisupervised Contrastive Learning for Bioactivity Prediction Using Cell Painting Image Data DOI Creative Commons

David Bushiri Pwesombo,

Carsten Jörn Beese, Christopher Schmied

et al.

Journal of Chemical Information and Modeling, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 6, 2025

Morphological profiling has recently demonstrated remarkable potential for identifying the biological activities of small molecules. Alongside fully supervised and self-supervised machine learning methods proposed bioactivity prediction from Cell Painting image data, we introduce here a semisupervised contrastive (SemiSupCon) approach. This approach combines strengths using annotations in leveraging large unannotated data sets with learning. SemiSupCon enhances downstream performance classifying MeSH pharmacological classifications PubChem, as well mode action target Drug Repurposing Hub across two publicly available sets. Notably, our effectively predicted several compounds, these findings were validated through literature searches. demonstrates that can potentially expedite exploration activity based on minimal human intervention.

Language: Английский

Citations

0
Cancer drug resistance as learning of signaling networks DOI
Dávid Keresztes, Márk Kerestély,

Levente Szarka

et al.

Biomedicine & Pharmacotherapy, Journal Year: 2025, Volume and Issue: 183, P. 117880 - 117880

Published: Jan. 29, 2025

Language: Английский

Citations

0
System level network data and models attack cancer drug resistance DOI Creative Commons
Márk Kerestély, Dávid Keresztes,

Levente Szarka

et al.

British Journal of Pharmacology, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 5, 2025

Drug resistance is responsible for >90% of cancer related deaths. Cancer drug a system level network phenomenon covering the entire cell. Small‐scale interactomes and signalling models guide directed development. Recently, proteome‐wide human interactome data have become available, which been extended by drug–target interactions, resistance‐inducing mutations, as well several resistance‐related multi‐omics datasets. System available examining therapy resistance, performing in silico clinical trials, conducting large, combination screens. interoperable reliable. These advances paved road building models.

Language: Английский

Citations

0
Optimizing Model Performance and Interpretability: Application to Biological Data Classification DOI Open Access
Zhenyu Huang,

Xuechen Mu,

Yangkun Cao

et al.

Genes, Journal Year: 2025, Volume and Issue: 16(3), P. 297 - 297

Published: Feb. 28, 2025

This study introduces a novel framework that simultaneously addresses the challenges of performance accuracy and result interpretability in transcriptomic-data-based classification. Background/objectives: In biological data classification, it is challenging to achieve both high at same time. presents address The goal select features, models, meta-voting classifier optimizes classification interpretability. Methods: consists four-step feature selection process: (1) identification metabolic pathways whose enzyme-gene expressions discriminate samples with different labels, aiding interpretability; (2) expression variance largely captured by first principal component gene matrix; (3) minimal sets genes, collective discerning power covers 95% pathway-based power; (4) introduction adversarial identify filter genes sensitive such samples. Additionally, are used optimal model, constructed based on optimized model results. Results: applied two cancer problems showed binary prediction was comparable full-gene F1-score differences between −5% 5%. ternary significantly better, ranging from −2% 12%, while also maintaining excellent selected genes. Conclusions: effectively integrates selection, sample handling, optimization, offering valuable tool for wide range problems. Its ability balance makes highly applicable field computational biology.

Language: Английский

Citations

0
Integrative multi-omics analysis and machine learning refine global histone modification features in prostate cancer DOI Creative Commons
Xiaofeng He, Qintao Ge, Weihao Zhao

et al.

Frontiers in Molecular Biosciences, Journal Year: 2025, Volume and Issue: 12

Published: March 12, 2025

Background Prostate cancer (PCa) is a major cause of cancer-related mortality in men, characterized by significant heterogeneity clinical behavior and treatment response. Histone modifications play key roles tumor progression resistance, but their regulatory effects PCa remain poorly understood. Methods We utilized integrative multi-omics analysis machine learning to explore histone modification-driven PCa. The Comprehensive Machine Learning Modification Score (CMLHMS) was developed classify into two distinct subtypes based on modification patterns. Single-cell RNA sequencing performed, drug sensitivity identified potential therapeutic vulnerabilities. Results High-CMLHMS tumors exhibited elevated activity, enriched proliferative metabolic pathways, were strongly associated with castration-resistant prostate (CRPC). Low-CMLHMS showed stress-adaptive immune-regulatory phenotypes. revealed differentiation trajectories related aggressiveness Drug that high-CMLHMS more responsive growth factor kinase inhibitors (e.g., PI3K, EGFR inhibitors), while low-CMLHMS demonstrated greater cytoskeletal DNA damage repair-targeting agents Paclitaxel, Gemcitabine). Conclusion CMLHMS model effectively stratifies unique biological characteristics. This study provides new insights suggests targets, contributing precision oncology strategies for advanced

Language: Английский

Citations

0
Structural and Functional Impacts of SARS-CoV-2 Spike Protein Mutations: Insights from Predictive Modeling and Analytics (Preprint) DOI Creative Commons

Edem K. Netsey,

Samuel M. Naandam,

Joseph Jr. Asante

et al.

Published: March 8, 2025

BACKGROUND The COVID-19 pandemic requires a deep understanding of SARS-CoV-2, particularly how mutations in the Spike Receptor Binding Domain (RBD) Chain E affect its structure and function. Current methods lack comprehensive analysis these at different structural levels. OBJECTIVE To analyze impact specific associated point (N501Y, L452R, N440K, K417N, E484A) on SARS-CoV-2 RBD function using predictive modeling, including graph-theoretic model, protein modeling techniques, molecular dynamics simulations. METHODS study employed multi-tiered framework to represent across three interconnected This model incorporated 19 top-level vertices, connected intermediate graphs based 6-angstrom proximity within protein's 3D structure. Graph-theoretic metrics were applied weigh vertices edges all also used Iterative Threading Assembly Refinement (I-TASSER) mutated sequences dynamic simulation (MD) tools evaluate changes folding stability compared wildtype. RESULTS Three distinct analytical approaches successfully identified functional (Chain E) resulting from mutations. novel detected notable changes, with N501Y L452R showing most pronounced effects conformation stability. K147N E484A demonstrated less significant impacts. Ab initio MD findings corroborated analysis. multi-level approach provided visualization mutation effects, deepening our their consequences. CONCLUSIONS advanced implications. multi-faceted characterized various mutations, identifying as having substantial have important implications for vaccine development, therapeutic design, variant monitoring. research underscores power combining multiple virology, contributing valuable knowledge ongoing efforts against providing future studies viral impacts

Language: Английский

Citations

0
Revolutionizing multi‐omics analysis with artificial intelligence and data processing DOI Open Access
Ali Yetgin

Quantitative Biology, Journal Year: 2025, Volume and Issue: 13(3)

Published: April 7, 2025

Language: Английский

Citations

0
Artificial intelligence-driven clinical decision support systems for early detection and precision therapy in oral cancer: a mini review DOI Creative Commons

Manoj Kumar Karuppan Perumal,

Remya Rajan Renuka,

Suresh Kumar

et al.

Frontiers in Oral Health, Journal Year: 2025, Volume and Issue: 6

Published: April 28, 2025

Oral cancer (OC) is a significant global health burden, with life-saving improvements in survival and outcomes being dependent on early diagnosis precise treatment planning. However, planning are predicated the synthesis of complicated information derived from clinical assessment, imaging, histopathology patient histories. Artificial intelligence-based decision support systems (AI-CDSS) provides viable solution that can be implemented via advanced methodologies for data analysis, better diagnostic prognostic evaluation. This review presents AI-CDSS as promising through comprehensive analysis. In addition, it examines current implementations facilitate OC detection, staging, personalized by processing multimodal machine learning, computer vision, natural language processing. These effectively interpret results, identify critical disease patterns (including stage, site, tumor dimensions, histopathologic grading, molecular profiles), construct profiles. approach allows reduction delays improved intervention outcomes. Moreover, also optimizes plans basis unique parameters, stages risk factors, providing therapy.

Language: Английский

Citations

0
Integrative Metabolome and Proteome Analysis of Cerebrospinal Fluid in Parkinson’s Disease DOI Open Access
Seok Gi Kim, Ji Su Hwang,

Nimisha Pradeep George

et al.

International Journal of Molecular Sciences, Journal Year: 2024, Volume and Issue: 25(21), P. 11406 - 11406

Published: Oct. 23, 2024

Parkinson's disease (PD) is a common neurodegenerative disorder characterized by the loss of dopaminergic neurons in substantia nigra. Recent studies have highlighted significant role cerebrospinal fluid (CSF) reflecting pathophysiological PD brain conditions analyzing components CSF. Based on published literature, we created single network with altered metabolites CSF patients PD. We analyzed biological functions related to transmembrane mitochondria, respiration neurodegeneration, and using bioinformatics tool. As proteome reflects phenotypes, collected data based papers, function showed similarities that metabolomic network. Then, integrated metabolome proteome. In silico predictions metabolomics proteomics neurodegeneration were predicted be activated. contrast, mitochondrial activity suppressed This review underscores importance omics analyses deciphering PD's complex biochemical networks underlying neurodegeneration.

Language: Английский

Citations

3