Interpretable deep learning in single-cell omics DOI Creative Commons
Manoj M Wagle, Siqu Long, Carissa Chen

и другие.

Bioinformatics, Год журнала: 2024, Номер 40(6)

Опубликована: Июнь 1, 2024

Abstract Motivation Single-cell omics technologies have enabled the quantification of molecular profiles in individual cells at an unparalleled resolution. Deep learning, a rapidly evolving sub-field machine has instilled significant interest single-cell research due to its remarkable success analysing heterogeneous high-dimensional data. Nevertheless, inherent multi-layer nonlinear architecture deep learning models often makes them ‘black boxes’ as reasoning behind predictions is unknown and not transparent user. This stimulated increasing body for addressing lack interpretability models, especially data analyses, where identification understanding regulators are crucial interpreting model directing downstream experimental validations. Results In this work, we introduce basics concept interpretable learning. followed by review recent applied various research. Lastly, highlight current limitations discuss potential future directions.

Язык: Английский

scGRN: a comprehensive single-cell gene regulatory network platform of human and mouse DOI Creative Commons
Xuemei Huang, Chao Song, Guorui Zhang

и другие.

Nucleic Acids Research, Год журнала: 2023, Номер 52(D1), С. D293 - D303

Опубликована: Окт. 27, 2023

Gene regulatory networks (GRNs) are interpretable graph models encompassing the interactions between transcription factors (TFs) and their downstream target genes. Making sense of topology dynamics GRNs is fundamental to interpreting mechanisms disease etiology translating corresponding findings into novel therapies. Recent advances in single-cell multi-omics techniques have prompted computational inference from transcriptomic epigenomic data at an unprecedented resolution. Here, we present scGRN (https://bio.liclab.net/scGRN/), a comprehensive gene network platform human mouse. The current version catalogs 237 051 cell type-specific (62 999 692 TF-target pairs), covering 160 tissues/cell lines 1324 samples. first resource documenting large-scale GRN information diverse mouse conditions inferred data. We implemented multiple online tools for effective analysis, including differential TF enrichment pathway analysis. also provided details about binding promoters, super-enhancers typical enhancers genes GRNs. Taken together, integrative useful searching, browsing, analyzing, visualizing downloading interest, enabling insight differences across conditions.

Язык: Английский

Процитировано

12
A bioinformatics tool for identifying intratumoral microbes from the ORIEN dataset DOI Creative Commons
Cankun Wang, Anjun Ma, Yingjie Li

и другие.

Cancer Research Communications, Год журнала: 2024, Номер 4(2), С. 293 - 302

Опубликована: Янв. 23, 2024

Abstract Evidence supports significant interactions among microbes, immune cells, and tumor cells in at least 10%–20% of human cancers, emphasizing the importance further investigating these complex relationships. However, implications significance tumor-related microbes remain largely unknown. Studies have demonstrated critical roles host cancer prevention treatment responses. Understanding between can drive diagnosis microbial therapeutics (bugs as drugs). Computational identification cancer-specific their associations is still challenging due to high dimensionality sparsity intratumoral microbiome data, which requires large datasets containing sufficient event observations identify relationships, within communities, heterogeneity composition, other confounding effects that lead spurious associations. To solve issues, we present a bioinformatics tool, graph attention (MEGA), most strongly associated with 12 types. We demonstrate its utility on dataset from consortium nine centers Oncology Research Information Exchange Network. This package has three unique features: species-sample relations are represented heterogeneous learned by network; it incorporates metabolic phylogenetic information reflect intricate relationships communities; provides multiple functionalities for association interpretations visualizations. analyzed 2,704 RNA sequencing samples MEGA interpreted tissue-resident signatures each effectively cancer-associated refine tumors. Significance: Studying high-throughput data because extremely sparse matrices, heterogeneity, likelihood contamination. new deep learning MEGA, organisms interact

Язык: Английский

Процитировано

4
A weighted two-stage sequence alignment framework to identify motifs from ChIP-exo data DOI Creative Commons
Yang Li, Yi‐Zhong Wang, Cankun Wang

и другие.

Patterns, Год журнала: 2024, Номер 5(3), С. 100927 - 100927

Опубликована: Фев. 2, 2024

In this study, we introduce TESA (weighted two-stage alignment), an innovative motif prediction tool that refines the identification of DNA-binding protein motifs, essential for deciphering transcriptional regulatory mechanisms. Unlike traditional algorithms rely solely on sequence data, integrates high-resolution chromatin immunoprecipitation (ChIP) signal, specifically from ChIP-exonuclease (ChIP-exo), by assigning weights to positions, thereby enhancing discovery. employs a nuanced approach combining binomial distribution model with graph model, further supported "bookend" improve accuracy predicting motifs varying lengths. Our evaluation, utilizing extensive compilation 90 prokaryotic ChIP-exo datasets proChIPdb and 167 H. sapiens datasets, compared TESA's performance against seven established tools. The results indicate improved precision in identification, suggesting its valuable contribution field genomic research.

Язык: Английский

Процитировано

4
Multi-view discriminative edge heterophily contrastive learning network for attributed graph anomaly detection DOI
Wangyu Jin, Huifang Ma, Yingyue Zhang

и другие.

Expert Systems with Applications, Год журнала: 2024, Номер 255, С. 124460 - 124460

Опубликована: Июнь 12, 2024

Язык: Английский

Процитировано

4
Interpretable deep learning in single-cell omics DOI Creative Commons
Manoj M Wagle, Siqu Long, Carissa Chen

и другие.

Bioinformatics, Год журнала: 2024, Номер 40(6)

Опубликована: Июнь 1, 2024

Abstract Motivation Single-cell omics technologies have enabled the quantification of molecular profiles in individual cells at an unparalleled resolution. Deep learning, a rapidly evolving sub-field machine has instilled significant interest single-cell research due to its remarkable success analysing heterogeneous high-dimensional data. Nevertheless, inherent multi-layer nonlinear architecture deep learning models often makes them ‘black boxes’ as reasoning behind predictions is unknown and not transparent user. This stimulated increasing body for addressing lack interpretability models, especially data analyses, where identification understanding regulators are crucial interpreting model directing downstream experimental validations. Results In this work, we introduce basics concept interpretable learning. followed by review recent applied various research. Lastly, highlight current limitations discuss potential future directions.

Язык: Английский

Процитировано

4