Deep transfer learning for clinical decision-making based on high-throughput data: comprehensive survey with benchmark results DOI
Muhammad Toseef, Olutomilayo Olayemi Petinrin, Fuzhou Wang

et al.

Briefings in Bioinformatics, Journal Year: 2023, Volume and Issue: 24(4)

Published: July 1, 2023

The rapid growth of omics-based data has revolutionized biomedical research and precision medicine, allowing machine learning models to be developed for cutting-edge performance. However, despite the wealth high-throughput available, performance these is hindered by lack sufficient training data, particularly in clinical (in vivo experiments). As a result, translating this knowledge into practice, such as predicting drug responses, remains challenging task. Transfer promising tool that bridges gap between domains transferring from source target domain. Researchers have proposed transfer predict outcomes leveraging pre-clinical (mouse, zebrafish), highlighting its vast potential. In work, we present comprehensive literature review deep methods health informatics decision-making, focusing on molecular data. Previous reviews mostly covered image-based works, while more detailed analysis papers. Furthermore, evaluated original studies based different evaluation settings across cross-validations, splits model architectures. result shows those great potential; sequencing state-of-the-art lead significant insights conclusions. Additionally, explored various datasets papers with statistics visualization.

Language: Английский

The Evolution of Single-Cell RNA Sequencing Technology and Application: Progress and Perspectives DOI Open Access
Shuo Wang,

Si-Tong Sun,

Xinyue Zhang

et al.

International Journal of Molecular Sciences, Journal Year: 2023, Volume and Issue: 24(3), P. 2943 - 2943

Published: Feb. 2, 2023

As an emerging sequencing technology, single-cell RNA (scRNA-Seq) has become a powerful tool for describing cell subpopulation classification and heterogeneity by achieving high-throughput multidimensional analysis of individual cells circumventing the shortcomings traditional detecting average transcript level populations. It been applied to life science medicine research fields such as tracking dynamic differentiation, revealing sensitive effector cells, key molecular events diseases. This review focuses on recent technological innovations in scRNA-Seq, highlighting latest results with scRNA-Seq core technology frontier areas embryology, histology, oncology, immunology. In addition, this outlines prospects its innovative application Chinese (TCM) discusses issues currently being addressed great potential exploring disease diagnostic targets uncovering drug therapeutic combination multiomics technologies.

Language: Английский

Citations

77

Probabilistic embedding, clustering, and alignment for integrating spatial transcriptomics data with PRECAST DOI Creative Commons
Wei Liu, Xu Liao, Ziye Luo

et al.

Nature Communications, Journal Year: 2023, Volume and Issue: 14(1)

Published: Jan. 18, 2023

Spatially resolved transcriptomics involves a set of emerging technologies that enable the transcriptomic profiling tissues with physical location expressions. Although variety methods have been developed for data integration, most them are single-cell RNA-seq datasets without consideration spatial information. Thus, can integrate from multiple tissue slides, possibly individuals, needed. Here, we present PRECAST, integration method complex batch effects and/or biological between slides. PRECAST unifies factor analysis simultaneously clustering and embedding alignment, while requiring only partially shared cell/domain clusters across datasets. Using both simulated four real datasets, show improved detection outstanding visualization, estimated aligned embeddings labels facilitate many downstream analyses. We demonstrate is computationally scalable applicable to different platforms.

Language: Английский

Citations

75

Deep learning applications in single-cell genomics and transcriptomics data analysis DOI Creative Commons
Nafiseh Erfanian, A. Ali Heydari, Adib Miraki Feriz

et al.

Biomedicine & Pharmacotherapy, Journal Year: 2023, Volume and Issue: 165, P. 115077 - 115077

Published: July 1, 2023

Traditional bulk sequencing methods are limited to measuring the average signal in a group of cells, potentially masking heterogeneity, and rare populations. The single-cell resolution, however, enhances our understanding complex biological systems diseases, such as cancer, immune system, chronic diseases. However, technologies generate massive amounts data that often high-dimensional, sparse, complex, thus making analysis with traditional computational approaches difficult unfeasible. To tackle these challenges, many turning deep learning (DL) potential alternatives conventional machine (ML) algorithms for studies. DL is branch ML capable extracting high-level features from raw inputs multiple stages. Compared ML, models have provided significant improvements across domains applications. In this work, we examine applications genomics, transcriptomics, spatial multi-omics integration, address whether techniques will prove be advantageous or if omics domain poses unique challenges. Through systematic literature review, found has not yet revolutionized most pressing challenges field. using shown promising results (in cases outperforming previous state-of-the-art models) preprocessing downstream analysis. Although developments generally been gradual, recent advances reveal can offer valuable resources fast-tracking advancing research single-cell.

Language: Английский

Citations

62

Inferring gene regulatory networks from single-cell multiome data using atlas-scale external data DOI Creative Commons
Qiuyue Yuan, Zhana Duren

Nature Biotechnology, Journal Year: 2024, Volume and Issue: unknown

Published: April 12, 2024

Abstract Existing methods for gene regulatory network (GRN) inference rely on expression data alone or lower resolution bulk data. Despite the recent integration of chromatin accessibility and RNA sequencing data, learning complex mechanisms from limited independent points still presents a daunting challenge. Here we present LINGER (Lifelong neural regulation), machine-learning method to infer GRNs single-cell paired incorporates atlas-scale external across diverse cellular contexts prior knowledge transcription factor motifs as manifold regularization. achieves fourfold sevenfold relative increase in accuracy over existing reveals landscape genome-wide association studies, enabling enhanced interpretation disease-associated variants genes. Following GRN reference multiome enables estimation activity solely leveraging abundance available identify driver regulators case-control studies.

Language: Английский

Citations

28

Online single-cell data integration through projecting heterogeneous datasets into a common cell-embedding space DOI Creative Commons
Lei Xiong, Tian Kang, Yuzhe Li

et al.

Nature Communications, Journal Year: 2022, Volume and Issue: 13(1)

Published: Oct. 17, 2022

Abstract Computational tools for integrative analyses of diverse single-cell experiments are facing formidable new challenges including dramatic increases in data scale, sample heterogeneity, and the need to informatively cross-reference with foundational datasets. Here, we present SCALEX, a deep-learning method that integrates by projecting cells into batch-invariant, common cell-embedding space truly online manner (i.e., without retraining model). SCALEX substantially outperforms iNMF other state-of-the-art non-online integration methods on benchmark datasets modalities, (e.g., RNA sequencing, scRNA-seq, assay transposase-accessible chromatin use scATAC-seq), especially partial overlaps, accurately aligning similar cell populations while retaining true biological differences. We showcase SCALEX’s advantages constructing continuously expandable atlases human, mouse, COVID-19 patients, each assembled from sources growing every data. The capacity superior performance makes particularly appropriate large-scale applications build upon previous scientific insights.

Language: Английский

Citations

67

scDART: integrating unmatched scRNA-seq and scATAC-seq data and learning cross-modality relationship simultaneously DOI Creative Commons
Ziqi Zhang,

Chengkai Yang,

Xiuwei Zhang

et al.

Genome biology, Journal Year: 2022, Volume and Issue: 23(1)

Published: June 27, 2022

Abstract It is a challenging task to integrate scRNA-seq and scATAC-seq data obtained from different batches. Existing methods tend use pre-defined gene activity matrix convert the into data. The often of low quality does not reflect dataset-specific relationship between two modalities. We propose scDART, deep learning framework that integrates learns cross-modalities relationships simultaneously. Specifically, design scDART allows it preserve cell trajectories in continuous populations can be applied trajectory inference on integrated

Language: Английский

Citations

54

MIRA: joint regulatory modeling of multimodal expression and chromatin accessibility in single cells DOI
Allen W. Lynch, Christina V. Theodoris, Henry W. Long

et al.

Nature Methods, Journal Year: 2022, Volume and Issue: 19(9), P. 1097 - 1108

Published: Sept. 1, 2022

Language: Английский

Citations

49

A unified computational framework for single-cell data integration with optimal transport DOI Creative Commons
Kai Cao,

Qiyu Gong,

Yiguang Hong

et al.

Nature Communications, Journal Year: 2022, Volume and Issue: 13(1)

Published: Dec. 1, 2022

Abstract Single-cell data integration can provide a comprehensive molecular view of cells. However, how to integrate heterogeneous single-cell multi-omics as well spatially resolved transcriptomic remains major challenge. Here we introduce uniPort, unified framework that combines coupled variational autoencoder (coupled-VAE) and minibatch unbalanced optimal transport (Minibatch-UOT). It leverages both highly variable common dataset-specific genes for handle the heterogeneity across datasets, it is scalable large-scale datasets. uniPort jointly embeds datasets into shared latent space. further construct reference atlas gene imputation Meanwhile, provides flexible label transfer deconvolute spatial using an plan, instead embedding We demonstrate capability by applying variety including transcriptomics, chromatin accessibility, data.

Language: Английский

Citations

47

Integration of spatial and single-cell data across modalities with weakly linked features DOI Creative Commons
Shuxiao Chen, Bokai Zhu, Sijia Huang

et al.

Nature Biotechnology, Journal Year: 2023, Volume and Issue: 42(7), P. 1096 - 1106

Published: Sept. 7, 2023

Abstract Although single-cell and spatial sequencing methods enable simultaneous measurement of more than one biological modality, no technology can capture all modalities within the same cell. For current data integration methods, feasibility cross-modal relies on existence highly correlated, a priori ‘linked’ features. We describe matching X-modality via fuzzy smoothed embedding (MaxFuse), method that, through iterative coembedding, smoothing cell matching, uses information in each modality to obtain high-quality even when features are weakly linked. MaxFuse is modality-agnostic demonstrates high robustness accuracy weak linkage scenario, achieving 20~70% relative improvement over existing under key evaluation metrics benchmarking datasets. A prototypical example proteomic with data. On two analyses this type, enabled consolidation proteomic, transcriptomic epigenomic at resolution tissue section.

Language: Английский

Citations

42

A global view of aging and Alzheimer’s pathogenesis-associated cell population dynamics and molecular signatures in human and mouse brains DOI Creative Commons

Andras Sziraki,

Ziyu Lu,

Jasper Lee

et al.

Nature Genetics, Journal Year: 2023, Volume and Issue: 55(12), P. 2104 - 2116

Published: Nov. 30, 2023

Abstract Conventional methods fall short in unraveling the dynamics of rare cell types related to aging and diseases. Here we introduce EasySci, an advanced single-cell combinatorial indexing strategy for exploring age-dependent cellular mammalian brain. Profiling approximately 1.5 million transcriptomes 400,000 chromatin accessibility profiles across diverse mouse brains, identified over 300 subtypes, uncovering their molecular characteristics spatial locations. This comprehensive view elucidates expanded or depleted upon aging. We also investigated cell-type-specific responses genetic alterations linked Alzheimer’s disease, identifying associated types. Additionally, by profiling 118,240 human brain transcriptomes, discerned cell- region-specific transcriptomic changes tied pathogenesis. In conclusion, this research offers a valuable resource probing both normal pathological

Language: Английский

Citations

39