scGraph2Vec: a deep generative model for gene embedding augmented by graph neural network and single-cell omics data DOI Creative Commons
Shiqi Lin, Peilin Jia

GigaScience, Journal Year: 2024, Volume and Issue: 13

Published: Jan. 1, 2024

Abstract Background Exploring the cellular processes of genes from aspects biological networks is great interest to understanding properties complex diseases and systems. Biological networks, such as protein–protein interaction gene regulatory provide insights into molecular basis often form functional clusters in different tissue disease contexts. Results We present scGraph2Vec, a deep learning framework for generating informative embeddings. scGraph2Vec extends variational graph autoencoder integrates single-cell datasets gene–gene networks. demonstrate that embeddings are biologically interpretable enable identification representing or tissue-specific processes. By comparing similar tools, we showed clearly distinguished aggregated more genes. can be widely applied diverse illustrated generated by infer disease-associated genome-wide association study data (e.g., COVID-19 Alzheimer's disease), identify additional driver lung adenocarcinoma, reveal responsible maintaining transitioning melanoma cell states. Conclusions not only reconstructs but also obtains latent representation implying their functions.

Language: Английский

Exploring the Molecular Terrain: A Survey of Analytical Methods for Biological Network Analysis DOI Open Access
Trong-The Nguyen, Thi-Kien Dao, Duc-Tinh Pham

et al.

Symmetry, Journal Year: 2024, Volume and Issue: 16(4), P. 462 - 462

Published: April 10, 2024

Biological systems, characterized by their complex interplay of symmetry and asymmetry, operate through intricate networks interacting molecules, weaving the elaborate tapestry life. The exploration these networks, aptly termed “molecular terrain”, is pivotal for unlocking mysteries biological processes spearheading development innovative therapeutic strategies. This review embarks on a comprehensive survey analytical methods employed in network analysis, focusing elucidating roles asymmetry within networks. By highlighting strengths, limitations, potential applications, we delve into reconstruction, topological analysis with an emphasis detection, examination dynamics, which together reveal nuanced balance between stable, symmetrical configurations dynamic, asymmetrical shifts that underpin functionality. equips researchers multifaceted toolbox designed to navigate decipher networks’ intricate, balanced landscape, thereby advancing our understanding manipulation systems. Through this detailed exploration, aim foster significant advancements paving way novel interventions deeper comprehension molecular underpinnings

Language: Английский

Citations

7

Integration of unpaired single cell omics data by deep transfer graph convolutional network DOI Creative Commons
Yulong Kan,

Y. Qi,

Zhongxiao Zhang

et al.

PLoS Computational Biology, Journal Year: 2025, Volume and Issue: 21(1), P. e1012625 - e1012625

Published: Jan. 16, 2025

The rapid advance of large-scale atlas-level single cell RNA sequences and single-cell chromatin accessibility data provide extraordinary avenues to broad deep insight into complex biological mechanism. Leveraging the datasets transfering labels from scRNA-seq scATAC-seq will empower exploration omics data. However, current label transfer methods have limited performance, largely due lower capable preserving fine-grained populations intrinsic or extrinsic heterogeneity between datasets. Here, we present a robust model based graph convolutional network, scTGCN, which achieves versatile performance in variation, while achieving integration hundreds thousands cells minutes with low memory consumption. We show that scTGCN is powerful mouse atlas multimodal generated APSA-seq CITE-seq. Thus, shows high accuracy effectively knowledge across different modalities.

Language: Английский

Citations

0

Graph neural networks for single-cell omics data: a review of approaches and applications DOI Creative Commons

Shiming Li,

Heyang Hua, Shengquan Chen

et al.

Briefings in Bioinformatics, Journal Year: 2025, Volume and Issue: 26(2)

Published: March 1, 2025

Abstract Rapid advancement of sequencing technologies now allows for the utilization precise signals at single-cell resolution in various omics studies. However, massive volume, ultra-high dimensionality, and high sparsity nature data have introduced substantial difficulties to traditional computational methods. The intricate non-Euclidean networks intracellular intercellular signaling molecules within datasets, coupled with complex, multimodal structures arising from multi-omics joint analysis, pose significant challenges conventional deep learning operations reliant on Euclidean geometries. Graph neural (GNNs) extended data, allowing cells their features datasets be modeled as nodes a graph structure. GNNs been successfully applied across broad range tasks analysis. In this survey, we systematically review 107 successful applications six variants tasks. We begin by outlining fundamental principles variants, followed systematic GNN-based models epigenomics, transcriptomics, spatial proteomics, multi-omics. each section dedicated specific type, summarized publicly available commonly utilized articles reviewed that section, totaling 77 datasets. Finally, summarize potential shortcomings current research explore directions future anticipate will serve guiding resource researchers deepen application omics.

Language: Английский

Citations

0

ZMGA: A ZINB-based multi-modal graph autoencoder enhancing topological consistency in single-cell clustering DOI

Jiaxi Yao,

Lin Li, Tongwen Xu

et al.

Biomedical Signal Processing and Control, Journal Year: 2024, Volume and Issue: 97, P. 106587 - 106587

Published: July 24, 2024

Language: Английский

Citations

1

TransGCN: a semi-supervised graph convolution network–based framework to infer protein translocations in spatio-temporal proteomics DOI Creative Commons
Bing Wang, Xiangzheng Zhang,

Xudong Han

et al.

Briefings in Bioinformatics, Journal Year: 2024, Volume and Issue: 25(2)

Published: Jan. 22, 2024

Abstract Protein subcellular localization (PSL) is very important in order to understand its functions, and movement between niches within cells plays fundamental roles biological process regulation. Mass spectrometry–based spatio-temporal proteomics technologies can help provide new insights of protein translocation, but bring the challenge identifying reliable translocation events due noise interference insufficient data mining. We propose a semi-supervised graph convolution network (GCN)–based framework termed TransGCN that infers from proteomics. Based on expanded multiple distance features joint representations proteins, utilizes GCN enable effective knowledge transfer proteins with known PSLs for predicting translocation. Our results demonstrate outperforms current state-of-the-art methods translocations, especially coping batch effects. It also exhibited excellent predictive accuracy PSL prediction. freely available GitHub at https://github.com/XuejiangGuo/TransGCN.

Language: Английский

Citations

0

scGraph2Vec: a deep generative model for gene embedding augmented by graph neural network and single-cell omics data DOI Creative Commons
Shiqi Lin, Peilin Jia

GigaScience, Journal Year: 2024, Volume and Issue: 13

Published: Jan. 1, 2024

Abstract Background Exploring the cellular processes of genes from aspects biological networks is great interest to understanding properties complex diseases and systems. Biological networks, such as protein–protein interaction gene regulatory provide insights into molecular basis often form functional clusters in different tissue disease contexts. Results We present scGraph2Vec, a deep learning framework for generating informative embeddings. scGraph2Vec extends variational graph autoencoder integrates single-cell datasets gene–gene networks. demonstrate that embeddings are biologically interpretable enable identification representing or tissue-specific processes. By comparing similar tools, we showed clearly distinguished aggregated more genes. can be widely applied diverse illustrated generated by infer disease-associated genome-wide association study data (e.g., COVID-19 Alzheimer's disease), identify additional driver lung adenocarcinoma, reveal responsible maintaining transitioning melanoma cell states. Conclusions not only reconstructs but also obtains latent representation implying their functions.

Language: Английский

Citations

0