BLEND: Probabilistic Cellular Deconvolution with Automated Reference Selection DOI Creative Commons
Penghui Huang, Manqi Cai, Chris McKennan

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Aug. 6, 2024

Cellular deconvolution aims to estimate cell type fractions from bulk transcriptomic and other omics data. Most existing methods fail account for the heterogeneity in type-specific (CTS) expression across samples, ignore discrepancies between CTS reference data, provide no guidance on selection or integration. To address these issues, we introduce BLEND, a hierarchical Bayesian method that leverages multiple datasets. BLEND learns most suitable references each sample by exploring convex hulls of employs "bag-of-words" representation count data deconvolution. speed up computation, an efficient EM algorithm parameter estimation. Notably, requires transformation, normalization, marker gene selection, quality evaluation. Benchmarking studies both simulated real human brain highlight BLEND's superior performance various scenarios. The analysis Alzheimer's disease illustrates application resource

Language: Английский

Benchmark of cellular deconvolution methods using a multi-assay dataset from postmortem human prefrontal cortex DOI Creative Commons
Louise A. Huuki-Myers, Kelsey D. Montgomery, Sang Ho Kwon

et al.

Genome biology, Journal Year: 2025, Volume and Issue: 26(1)

Published: April 7, 2025

Language: Английский

Citations

3

Transcriptomic characterization of human lateral septum neurons reveals conserved and divergent marker genes across species DOI Creative Commons
Robert A. Phillips, Seyun Oh, Svitlana V. Bach

et al.

iScience, Journal Year: 2025, Volume and Issue: 28(2), P. 111820 - 111820

Published: Jan. 18, 2025

The lateral septum (LS) is a midline, subcortical structure that critical regulator of social behaviors. Mouse studies have identified molecularly distinct neuronal populations within the LS, which control specific facets behavior. Despite its known molecular heterogeneity in mouse and role regulating behavior, comprehensive profiling human LS has not been performed. Here, we conducted single-nucleus RNA sequencing (snRNA-seq) to generate transcriptomic profiles compared recently collected snRNA-seq datasets. Our analyses TRPC4 as conserved marker while FREM2 enriched only LS. We also identify cell type marked by OPRM1, gene encoding μ-opioid receptor. Together, these results highlight transcriptional robust genes for

Language: Английский

Citations

2

An integrated single-nucleus and spatial transcriptomics atlas reveals the molecular landscape of the human hippocampus DOI Creative Commons
Erik D. Nelson, Madhavi Tippani, Anthony D. Ramnauth

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: April 28, 2024

Abstract The hippocampus contains many unique cell types, which serve the structure’s specialized functions, including learning, memory and cognition. These cells have distinct spatial topography, morphology, physiology, connectivity, highlighting need for transcriptome-wide profiling strategies that retain cytoarchitectural organization. Here, we generated spatially-resolved transcriptomics (SRT) single-nucleus RNA-sequencing (snRNA-seq) data from adjacent tissue sections of anterior human across ten adult neurotypical donors. We defined molecular profiles hippocampal types domains. Using non-negative matrix factorization transfer integrated these to define gene expression patterns within snRNA-seq infer in SRT data. With this approach, leveraged existing rodent datasets feature information on circuit connectivity neural activity induction make predictions about axonal projection targets likelihood ensemble recruitment spatially-defined cellular populations hippocampus. Finally, genome-wide association studies with transcriptomic identify enrichment genetic components neurodevelopmental, neuropsychiatric, neurodegenerative disorders domains, To comprehensive atlas accessible scientific community, both raw processed are freely available, through interactive web applications.

Language: Английский

Citations

8

SMART: spatial transcriptomics deconvolution using marker-gene-assisted topic model DOI Creative Commons
Chen Xi Yang, Don D. Sin, Raymond T. Ng

et al.

Genome biology, Journal Year: 2024, Volume and Issue: 25(1)

Published: Dec. 2, 2024

Abstract While spatial transcriptomics offer valuable insights into gene expression patterns within the context of tissue, many technologies do not have a single-cell resolution. Here, we present SMART, marker gene-assisted deconvolution method that simultaneously infers cell type-specific profile and cellular composition at each spot. Using multiple datasets, show SMART outperforms existing methods in realistic settings. It also provides two-stage approach to enhance its performance on subtypes. The covariate model enables identification differentially expressed genes across conditions, elucidating biological changes single-cell-type

Language: Английский

Citations

3

lute: estimating the cell composition of heterogeneous tissue with varying cell sizes using gene expression DOI Creative Commons
Sean K. Maden, Louise A. Huuki-Myers, Sang Ho Kwon

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: April 6, 2024

Relative cell type fraction estimates in bulk RNA-sequencing data are important to control for composition differences across heterogenous tissue samples. Current computational tools estimate relative RNA abundances rather than proportions tissues with varying sizes, leading biased estimates. We present lute, a tool accurately deconvolute types sizes. Our software wraps existing deconvolution algorithms standardized framework. Using simulated and real datasets, we demonstrate how lute adjusts sizes improve the accuracy of composition. Software is available from https://bioconductor.org/packages/lute.

Language: Английский

Citations

2

BLEND: Probabilistic Cellular Deconvolution with Automated Reference Selection DOI Creative Commons
Penghui Huang, Manqi Cai, Chris McKennan

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Aug. 6, 2024

Cellular deconvolution aims to estimate cell type fractions from bulk transcriptomic and other omics data. Most existing methods fail account for the heterogeneity in type-specific (CTS) expression across samples, ignore discrepancies between CTS reference data, provide no guidance on selection or integration. To address these issues, we introduce BLEND, a hierarchical Bayesian method that leverages multiple datasets. BLEND learns most suitable references each sample by exploring convex hulls of employs "bag-of-words" representation count data deconvolution. speed up computation, an efficient EM algorithm parameter estimation. Notably, requires transformation, normalization, marker gene selection, quality evaluation. Benchmarking studies both simulated real human brain highlight BLEND's superior performance various scenarios. The analysis Alzheimer's disease illustrates application resource

Language: Английский

Citations

0