CellDART Transcriptomic Data

DownLoad

Summary

Elucidating the cellular constituents within a tissue's spatial context through the examination of genome-wide spatially resolved transcriptomic information is paramount. The study pioneered a technique, CellDART, which gauges cell distributions delineated by single-cell data, employing domain adaptation of neural networks for spatial human lung tissue mapping. By translating the neural network predicting cellular proportions in pseudospots, virtual cell mixtures derived from single-cell information, the methodology disentangles cell types within spatially barcoded regions. Applied initially to mouse brains and human dorsolateral prefrontal cortex tissue, CellDART identified layer-specific cell type distributions, displaying superior stability, accuracy, and rapid execution compared to alternative computational strategies for determining excitatory neuron locations. Capable of decomposing cellular proportions in mouse hippocampus Slide-seq data, CellDART also revealed predominant cell types delineated by the human lung cell atlas within lung tissue compartments, aligning with known prevalent cell classifications. Ultimately, CellDART may illuminate spatial cellular heterogeneity and intimate intercellular interactions across diverse tissues.

Research Criteria

The research idea of this article is to develop a method called CellDART that estimates the spatial distribution of cells defined by single-cell level data using domain adaptation of neural networks and apply it to the spatial mapping of human lung tissue.

Fig.1 Experimental design. (Bae, 2022)

Sample Type

Human lung tissues.

Result—Decomposition of spatial cell distribution with CellDART in human and mouse brain data

Examination of single-nucleus and spatially resolved transcriptomic specimens procured from human dorsolateral prefrontal cortex and rodent cerebrum enabled researchers to delineate 33 and 29 discrete cell groupings, respectively, utilizing t-SNE plots. Pseudospots, generated via stochastic sampling of a predetermined cell quantity (k=8), facilitated neural network implementation to deconstruct pseudospots and instruct a domain classifier in discerning authentic spatially resolved transcriptomes from pseudospots. CellDART successfully forecasted laminar compositions of excitatory neurons in both human and murine cerebra, revealing spatially restricted patterns within distinct cortical strata. Moreover, the investigation underscored domain adaptation's import in enhancing model efficacy by juxtaposing CellDART's precision with a model devoid of domain adaptation, NN_wo_da. Predominantly, CellDART surpassed NN_wo_da, accentuating the indispensability of domain adaptation for optimal results. This trailblazing methodology paves the way for holistic explication of tissue organization at the cellular and tissue tiers within individual biological samples. In aggregate, CellDART manifested superior stability and precision, coupled with expeditious runtimes, compared to alternative methodologies predicting excitatory neuron spatial localization.

Fig.2 CellDART analysis in human and mouse brain tissues. (Bae, 2022)

Result—Discovery of spatial heterogeneity of human lung tissue with CellDART

In a pioneering endeavor, scholars scrutinized non-malignant human lung specimens procured from a cancer-afflicted individual, unveiling 57 distinct cell groupings delineated by discrete genetic expression profiles within the human pulmonary cell atlas, as illustrated by t-SNE plots. Through the application of CellDART, experts adroitly apportioned tissue sections into septenary classifications and quintet cell type categories, informed by antecedent inquiries. Individual cell types evinced singular dissemination trajectories throughout partitioned tissue domains, with certain varieties displaying conspicuously divergent cellular proportions across regions. Employing CellDART facilitated the precise demarcation of the spatiotemporal arrangement of heterogeneous cells in healthy pulmonary tissue, proffering insight into cellular orchestration and tissue functionality. This innovative approach ultimately enables thorough explication of tissue structure at the cellular and tissue echelons.

Fig.3 Application of CellDART in human lung data to decipher the tissue microenvironment. (Bae, 2022)

Creative Biolabs' Service

Single Cell Spatial Gene Expression Service

Creative Biolabs provides a comprehensive range of customized, high-quality services in single cell spatial gene expression service to help researchers map the whole transcriptome with morphological context in FFPE or fresh-frozen tissues to discover novel insights into normal development, disease pathology, and clinical translational research.

Learn more

Creative Biolabs presents an exclusive single cell spatial gene expression facility that assists scientists in mapping cell transcriptomes in tissues. This provision proves to be extremely valuable for examining healthy tissue development, comprehending the origin of diseases, and executing clinical translational research. Our adept panel of professionals can personalize the service to correspond to the unique requirements and objectives of every investigator.

For any information, please contact us.

Reference

1. Bae, S.; et al. CellDART: cell type inference by domain adaptation of single-cell and spatial transcriptomic data. Nucleic Acids Research. 2022, 50(10): e57-e57.