DMOIT: denoised multi-omics integration approach based on transformer multi-head self-attention mechanism DOI Creative Commons
Zhe Liu, Taesung Park

Frontiers in Genetics, Journal Year: 2024, Volume and Issue: 15

Published: Dec. 10, 2024

Multi-omics data integration has become increasingly crucial for a deeper understanding of the complexity biological systems. However, effectively integrating and analyzing multi-omics remains challenging due to their heterogeneity high dimensionality. Existing methods often struggle with noise, redundant features, complex interactions between different omics layers, leading suboptimal performance. Additionally, they face difficulties in adequately capturing intra-omics simplistic concatenation techiniques, risk losing critical inter-omics interaction information when using hierarchical attention layers. To address these challenges, we propose novel Denoised Multi-Omics Integration approach that leverages Transformer multi-head self-attention mechanism (DMOIT). DMOIT consists three key modules: generative adversarial imputation network handling missing values, sampling-based robust feature selection module reduce noise (MHSA) based extractor noval architecture enchance capture. We validated model porformance cancer datasets from Cancer Genome Atlas (TCGA), conducting two tasks: survival time classification across types estrogen receptor status breast cancer. Our results show outperforms traditional machine learning state-of-the-art method MoGCN terms accuracy weighted F1 score. Furthermore, compared various alternative MHSA-based architectures further validate our approach. consistently models combinations. The strong performance robustness demonstrate its potential as valuable tool applications.

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

1
The ROSMAP project: aging and neurodegenerative diseases through omic sciences DOI Creative Commons

Alejandra P. Pérez-González,

Aidee Lashmi García-Kroepfly,

Keila Adonai Pérez-Fuentes

et al.

Frontiers in Neuroinformatics, Journal Year: 2024, Volume and Issue: 18

Published: Sept. 16, 2024

The Religious Order Study and Memory Aging Project (ROSMAP) is an initiative that integrates two longitudinal cohort studies, which have been collecting clinicopathological molecular data since the early 1990s. This extensive dataset includes a wide array of omic data, revealing complex interactions between levels in neurodegenerative diseases (ND) aging. Neurodegenerative are frequently associated with morbidity cognitive decline older adults. Omics research, conjunction clinical variables, crucial for advancing our understanding diagnosis treatment diseases. summary reviews omics research—encompassing genomics, transcriptomics, proteomics, metabolomics, epigenomics, multiomics—conducted through ROSMAP study. It highlights significant advancements mechanisms underlying diseases, particular focus on Alzheimer's disease.

Language: Английский

Citations

7
Integrative approach of omics and imaging data to discover new insights for understanding brain diseases DOI Creative Commons
Jong Hyuk Yoon,

Hagyeong Lee,

Dayoung Kwon

et al.

Brain Communications, Journal Year: 2024, Volume and Issue: 6(4)

Published: Jan. 1, 2024

Abstract Treatments that can completely resolve brain diseases have yet to be discovered. Omics is a novel technology allows researchers understand the molecular pathways underlying diseases. Multiple omics, including genomics, transcriptomics and proteomics, imaging technologies, such as MRI, PET EEG, contributed disease-related therapeutic target detection. However, new treatment discovery remains challenging. We focused on establishing multi-molecular maps using an integrative approach of omics provide insights into disease diagnosis treatment. This requires precise data collection processing normalization. Incorporating map with advanced technologies through artificial intelligence will help establish system for regulation at level.

Language: Английский

Citations

4
A deep contrastive multi-modal encoder for multi-omics data integration and analysis DOI
Ma Yinghua, Ahmad Khan, Yang Heng

et al.

Information Sciences, Journal Year: 2025, Volume and Issue: 700, P. 121864 - 121864

Published: Jan. 7, 2025

Language: Английский

Citations

0
Multimodal data integration in early-stage breast cancer DOI Open Access
Arnau Llinas-Bertran,

Maria Butjosa-Espín,

Vittoria Barberi

et al.

The Breast, Journal Year: 2025, Volume and Issue: unknown, P. 103892 - 103892

Published: Jan. 1, 2025

Language: Английский

Citations

0
Network-based multi-omics integrative analysis methods in drug discovery: a systematic review DOI Creative Commons
Wei Jiang, Weicai Ye, Xiao-Ming Tan

et al.

BioData Mining, Journal Year: 2025, Volume and Issue: 18(1)

Published: March 28, 2025

The integration of multi-omics data from diverse high-throughput technologies has revolutionized drug discovery. While various network-based methods have been developed to integrate data, systematic evaluation and comparison these remain challenging. This review aims analyze approaches for evaluate their applications in We conducted a comprehensive literature (2015-2024) on discovery, categorized into four primary types: network propagation/diffusion, similarity-based approaches, graph neural networks, inference models. also discussed the three scenario including target identification, response prediction, repurposing, finally evaluated performance by highlighting advantages limitations specific applications. shown promise challenges computational scalability, integration, biological interpretation. Future developments should focus incorporating temporal spatial dynamics, improving model interpretability, establishing standardized frameworks.

Language: Английский

Citations

0
Advanced hybrid deep learning model for enhanced evaluation of osteosarcoma histopathology images DOI Creative Commons
Arezoo Borji, Gernot Kronreif,

Bernhard Angermayr

et al.

Frontiers in Medicine, Journal Year: 2025, Volume and Issue: 12

Published: April 16, 2025

Recent advances in machine learning are transforming medical image analysis, particularly cancer detection and classification. Techniques such as deep learning, especially convolutional neural networks (CNNs) vision transformers (ViTs), now enabling the precise analysis of complex histopathological images, automating detection, enhancing classification accuracy across various types. This study focuses on osteosarcoma (OS), most common bone children adolescents, which affects long bones arms legs. Early accurate OS is essential for improving patient outcomes reducing mortality. However, increasing prevalence demand personalized treatments create challenges achieving diagnoses customized therapies. We propose a novel hybrid model that combines (CNN) (ViT) to improve diagnostic using hematoxylin eosin (H&E) stained images. The CNN extracts local features, while ViT captures global patterns from These features combined classified Multi-Layer Perceptron (MLP) into four categories: non-tumor (NT), non-viable tumor (NVT), viable (VT), ratio (NVR). Using Cancer Imaging Archive (TCIA) dataset, achieved an 99.08%, precision 99.10%, recall 99.28%, F1-score 99.23%. first successful four-class this setting new benchmark research offering promising potential future advancements.

Language: Английский

Citations

0
A Hybrid Deep Learning Approach for Bearing Fault Diagnosis Using Continuous Wavelet Transform and Attention-Enhanced Spatiotemporal Feature Extraction DOI Creative Commons
Muhammad Siddique,

Faisal Saleem,

Muhammad Umar

et al.

Sensors, Journal Year: 2025, Volume and Issue: 25(9), P. 2712 - 2712

Published: April 25, 2025

This study presents a hybrid deep learning approach for bearing fault diagnosis that integrates continuous wavelet transform (CWT) with an attention-enhanced spatiotemporal feature extraction framework. The model combines time-frequency domain analysis using CWT classification architecture comprising multi-head self-attention (MHSA), bidirectional long short-term memory (BiLSTM), and 1D convolutional residual network (1D conv ResNet). effectively captures both spatial temporal dependencies, enhances noise resilience, extracts discriminative features from nonstationary nonlinear vibration signals. is initially trained on controlled laboratory dataset further validated real artificial subsets of the Paderborn dataset, demonstrating strong generalization across diverse conditions. t-SNE visualizations confirm clear separability between categories, supporting model’s capability precise reliable potential real-time predictive maintenance in complex industrial environments.

Language: Английский

Citations

0
LASSO–MOGAT: a multi-omics graph attention framework for cancer classification DOI Creative Commons

Fadi Alharbi,

Aleksandar Vakanski,

Murtada K. Elbashir

et al.

Academia Biology, Journal Year: 2024, Volume and Issue: 2(3)

Published: Aug. 30, 2024

The application of machine learning methods to analyze changes in gene expression patterns has recently emerged as a powerful approach cancer research, enhancing our understanding the molecular mechanisms underpinning development and progression. Combining data with other types omics been reported by numerous works improve classification outcomes. Despite these advances, effectively integrating high-dimensional multi-omics capturing complex relationships across different biological layers remains challenging. This paper introduces LASSO-MOGAT (LASSO-Multi-Omics Gated ATtention), novel graph-based deep framework that integrates messenger RNA, microRNA, DNA methylation classify 31 types. Utilizing differential analysis LIMMA LASSO regression for feature selection, leveraging Graph Attention Networks (GATs) incorporate protein-protein interaction (PPI) networks, captures intricate within data. Experimental validation using five-fold cross-validation demonstrates method's precision, reliability, capacity providing comprehensive insights into mechanisms. computation attention coefficients edges graph proposed graph-attention architecture based on interactions proved beneficial identifying synergies classification.

Language: Английский

Citations

3
Fault diagnosis method for lithium-ion batteries based on the combination of voltage prediction and Z-score DOI
Liao Li, Xunbo Li, Yang Da

et al.

International Journal of Green Energy, Journal Year: 2024, Volume and Issue: 21(14), P. 3270 - 3287

Published: July 9, 2024

Language: Английский

Citations

1