scMoMaT jointly performs single cell mosaic integration and multi-modal bio-marker detection DOI Creative Commons
Ziqi Zhang, Haoran Sun, Ragunathan Mariappan

et al.

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

Published: Jan. 24, 2023

Single cell data integration methods aim to integrate cells across batches and modalities, tasks can be categorized into horizontal, vertical, diagonal, mosaic integration, where is the most general challenging case with few developed. We propose scMoMaT, a method that able single multi-omics under scenario using matrix tri-factorization. During scMoMaT also uncover cluster specific bio-markers modalities. These multi-modal are used interpret annotate clusters types. Moreover, unequal type compositions. Applying multiple real simulated datasets demonstrated these features of showed has superior performance compared existing methods. Specifically, we show integrated embedding combined learned lead annotations higher quality or resolution their original annotations.

Language: Английский

Best practices for single-cell analysis across modalities DOI Open Access
Lukas Heumos, Anna C. Schaar, Christopher Lance

et al.

Nature Reviews Genetics, Journal Year: 2023, Volume and Issue: 24(8), P. 550 - 572

Published: March 31, 2023

Language: Английский

Citations

535
A Review of Single-Cell RNA-Seq Annotation, Integration, and Cell–Cell Communication DOI Creative Commons
Changde Cheng, Wenan Chen, Hongjian Jin

et al.

Cells, Journal Year: 2023, Volume and Issue: 12(15), P. 1970 - 1970

Published: July 30, 2023

Single-cell RNA sequencing (scRNA-seq) has emerged as a powerful tool for investigating cellular biology at an unprecedented resolution, enabling the characterization of heterogeneity, identification rare but significant cell types, and exploration cell-cell communications interactions. Its broad applications span both basic clinical research domains. In this comprehensive review, we survey current landscape scRNA-seq analysis methods tools, focusing on count modeling, cell-type annotation, data integration, including spatial transcriptomics, inference communication. We review challenges encountered in analysis, issues sparsity or low expression, reliability assumptions discuss potential impact suboptimal clustering differential expression tools downstream analyses, particularly identifying subpopulations. Finally, recent advancements future directions enhancing analysis. Specifically, highlight development novel annotating single-cell data, integrating interpreting multimodal datasets covering epigenomics, proteomics, inferring communication networks. By elucidating latest progress innovation, provide overview rapidly advancing field

Language: Английский

Citations

72
Integrated multimodal cell atlas of Alzheimer’s disease DOI Creative Commons
Mariano I. Gabitto, Kyle J. Travaglini, Victoria M. Rachleff

et al.

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

Published: Oct. 14, 2024

Alzheimer's disease (AD) is the leading cause of dementia in older adults. Although AD progression characterized by stereotyped accumulation proteinopathies, affected cellular populations remain understudied. Here we use multiomics, spatial genomics and reference atlases from BRAIN Initiative to study middle temporal gyrus cell types 84 donors with varying pathologies. This cohort includes 33 male 51 female donors, an average age at time death 88 years. We used quantitative neuropathology place along a pseudoprogression score. Pseudoprogression analysis revealed two phases: early phase slow increase pathology, presence inflammatory microglia, reactive astrocytes, loss somatostatin

Language: Английский

Citations

61
CellRank 2: unified fate mapping in multiview single-cell data DOI Creative Commons
Philipp Weiler, Marius Lange, Michal Klein

et al.

Nature Methods, Journal Year: 2024, Volume and Issue: 21(7), P. 1196 - 1205

Published: June 13, 2024

Abstract Single-cell RNA sequencing allows us to model cellular state dynamics and fate decisions using expression similarity or velocity reconstruct state-change trajectories; however, trajectory inference does not incorporate valuable time point information utilize additional modalities, whereas methods that address these different data views cannot be combined do scale. Here we present CellRank 2, a versatile scalable framework study multiview single-cell of up millions cells in unified fashion. 2 consistently recovers terminal states probabilities across modalities human hematopoiesis endodermal development. Our also combining transitions within experimental points, feature use recover genes promoting medullary thymic epithelial cell formation during pharyngeal endoderm Moreover, enable estimating cell-specific transcription degradation rates from metabolic-labeling data, which apply an intestinal organoid system delineate differentiation trajectories pinpoint regulatory strategies.

Language: Английский

Citations

33
The future of rapid and automated single-cell data analysis using reference mapping DOI Creative Commons
Mohammad Lotfollahi, Yuhan Hao, Fabian J. Theis

et al.

Cell, Journal Year: 2024, Volume and Issue: 187(10), P. 2343 - 2358

Published: May 1, 2024

As the number of single-cell datasets continues to grow rapidly, workflows that map new data well-curated reference atlases offer enormous promise for biological community. In this perspective, we discuss key computational challenges and opportunities reference-mapping algorithms. We how mapping algorithms will enable integration diverse across disease states, molecular modalities, genetic perturbations, species eventually replace manual laborious unsupervised clustering pipelines.

Language: Английский

Citations

28
Deciphering spatial domains from spatial multi-omics with SpatialGlue DOI Creative Commons
Yahui Long, Kok Siong Ang, Raman Sethi

et al.

Nature Methods, Journal Year: 2024, Volume and Issue: 21(9), P. 1658 - 1667

Published: June 21, 2024

Advances in spatial omics technologies now allow multiple types of data to be acquired from the same tissue slice. To realize full potential such data, we need spatially informed methods for integration. Here, introduce SpatialGlue, a graph neural network model with dual-attention mechanism that deciphers domains by intra-omics integration location and measurement followed cross-omics We demonstrated SpatialGlue on different using technologies, including epigenome-transcriptome transcriptome-proteome modalities. Compared other methods, captured more anatomical details accurately resolved as cortex layers brain. Our method also identified cell like spleen macrophage subsets located at three zones were not available original annotations. scales well size can used integrate multi-omics analysis tool combines information complementary modalities obtain holistic view cellular properties.

Language: Английский

Citations

27
Mosaic integration and knowledge transfer of single-cell multimodal data with MIDAS DOI Creative Commons
眞智子 平賀, Shuofeng Hu, Yaowen Chen

et al.

Nature Biotechnology, Journal Year: 2024, Volume and Issue: 42(10), P. 1594 - 1605

Published: Jan. 23, 2024

Abstract Integrating single-cell datasets produced by multiple omics technologies is essential for defining cellular heterogeneity. Mosaic integration, in which different share only some of the measured modalities, poses major challenges, particularly regarding modality alignment and batch effect removal. Here, we present a deep probabilistic framework mosaic integration knowledge transfer (MIDAS) multimodal data. MIDAS simultaneously achieves dimensionality reduction, imputation correction data using self-supervised information-theoretic latent disentanglement. We demonstrate its superiority to 19 other methods reliability evaluating performance trimodal tasks. also constructed atlas human peripheral blood mononuclear cells tailored learning reciprocal reference mapping schemes enable flexible accurate from new Applications pseudotime analysis cross-tissue on bone marrow versatility MIDAS. available at https://github.com/labomics/midas .

Language: Английский

Citations

24
Deep learning in spatially resolved transcriptomics: a comprehensive technical view DOI Creative Commons

Roxana Zahedi,

Mohammad Reza Eftekhariyan Ghamsari, Ahmadreza Argha

et al.

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

Published: Jan. 22, 2024

Abstract Spatially resolved transcriptomics (SRT) is a pioneering method for simultaneously studying morphological contexts and gene expression at single-cell precision. Data emerging from SRT are multifaceted, presenting researchers with intricate matrices, precise spatial details comprehensive histology visuals. Such rich datasets, unfortunately, render many conventional methods like traditional machine learning statistical models ineffective. The unique challenges posed by the specialized nature of data have led scientific community to explore more sophisticated analytical avenues. Recent trends indicate an increasing reliance on deep algorithms, especially in areas such as clustering, identification spatially variable genes alignment tasks. In this manuscript, we provide rigorous critique these advanced methodologies, probing into their merits, limitations avenues further refinement. Our in-depth analysis underscores that while recent innovations tailored been promising, there remains substantial potential enhancement. A crucial area demands attention development can incorporate biological nuances, phylogeny-aware processing or minuscule image segments. Furthermore, addressing elimination batch effects, perfecting normalization techniques countering overdispersion zero inflation patterns seen pivotal. To support broader endeavors, meticulously assembled directory readily accessible databases, hoping serve foundation future research initiatives.

Language: Английский

Citations

19
AI-driven multi-omics integration for multi-scale predictive modeling of genotype-environment-phenotype relationships DOI Creative Commons
You Wu, Lei Xie

Computational and Structural Biotechnology Journal, Journal Year: 2025, Volume and Issue: 27, P. 265 - 277

Published: Jan. 1, 2025

Despite the wealth of single-cell multi-omics data, it remains challenging to predict consequences novel genetic and chemical perturbations in human body. It requires knowledge molecular interactions at all biological levels, encompassing disease models humans. Current machine learning methods primarily establish statistical correlations between genotypes phenotypes but struggle identify physiologically significant causal factors, limiting their predictive power. Key challenges modeling include scarcity labeled generalization across different domains, disentangling causation from correlation. In light recent advances data integration, we propose a new artificial intelligence (AI)-powered biology-inspired multi-scale framework tackle these issues. This will integrate organism hierarchies, species genotype-environment-phenotype relationships under various conditions. AI inspired by biology may targets, biomarkers, pharmaceutical agents, personalized medicines for presently unmet medical needs.

Language: Английский

Citations

7
Mapping cells through time and space with moscot DOI Creative Commons
Dominik Klein, Giovanni Palla, Marius Lange

et al.

Nature, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 22, 2025

Abstract Single-cell genomic technologies enable the multimodal profiling of millions cells across temporal and spatial dimensions. However, experimental limitations hinder comprehensive measurement under native dynamics in their tissue niche. Optimal transport has emerged as a powerful tool to address these constraints facilitated recovery original cellular context 1–4 . Yet, most optimal applications are unable incorporate information or scale single-cell atlases. Here we introduce multi-omics (moscot), scalable framework for genomics that supports multimodality all applications. We demonstrate capability moscot efficiently reconstruct developmental trajectories 1.7 million from mouse embryos 20 time points. To illustrate space, enrich transcriptomic datasets by mapping profiles liver sample align multiple coronal sections brain. present moscot.spatiotemporal, an approach leverages gene-expression data both dimensions uncover spatiotemporal embryogenesis. also resolve endocrine-lineage relationships delta epsilon previously unpublished mouse, time-resolved pancreas development dataset using paired measurements gene expression chromatin accessibility. Our findings confirmed through validation NEUROD2 regulator progenitor model human induced pluripotent stem cell islet differentiation. Moscot is available open-source software, accompanied extensive documentation.

Language: Английский

Citations

7