Expression Atlas update: from tissues to single cells.
Papatheodorou I, Moreno P, Manning J, Fuentes AM, George N, Fexova S, Fonseca NA, Füllgrabe A, Green M, Huang N, Huerta L, Iqbal H, Jianu M, Mohammed S, Zhao L, Jarnuczak AF, Jupp S, Marioni J, Meyer K, Petryszak R, Prada Medina CA, Talavera-López C, Teichmann S, Vizcaino JA, Brazma A. · 2020
Read Expression Atlas update: from tissues to single cells. by Papatheodorou I, Moreno P, Manning J, Fuentes AM, George N, Fexova S, Fonseca NA, Füllgrabe A, Green M, Huang N, Huerta L, Iqbal H, Jianu M, Mohammed S, Zhao L, Jarnuczak AF, Jupp S, Marioni J, Meyer K, Petryszak R, Prada Medina CA, Talavera-López C, Teichmann S, Vizcaino JA, Brazma A. (2020) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc7145605.txt.
Introduction
Papatheodorou and co-authors update Expression Atlas, the EMBL-EBI resource for gene and protein expression, in Nucleic Acids Research. The atlas gathers measurements of how active each gene is across species, tissues, cells, diseases, and experimental conditions, and serves them openly. Its newest branch covers single-cell data.
Single Cell Expression Atlas launched in 2018 and already held 123 studies from 12 species at the time of writing. Users can search a gene across species to find where it is expressed or where it marks a cell type. Each study offers visual maps of cells coloured by cluster or feature, plus downloadable count tables and metadata. The update marks the atlas's move from tissue averages to single-cell resolution.
Catalog record
| CREATOR | Papatheodorou I, Moreno P, Manning J, Fuentes AM, George N, Fexova S, Fonseca NA, Füllgrabe A, Green M, Huang N, Huerta L, Iqbal H, Jianu M, Mohammed S, Zhao L, Jarnuczak AF, Jupp S, Marioni J, Meyer K, Petryszak R, Prada Medina CA, Talavera-López C, Teichmann S, Vizcaino JA, Brazma A. |
|---|---|
| YEAR | 2020 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 25 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Expression Atlas update: from tissues to single cells By Irene Papatheodorou, Pablo Moreno, Jonathan Manning, Alfonso Muñoz-Pomer Fuentes, Nancy George, Silvie Fexova, Nuno A Fonseca, Anja Füllgrabe, Matthew Green, Ni Huang, Laura Huerta, Haider Iqbal Published in: Nucleic Acids Research DOI: 10.1093/nar/gkz947 ABSTRACT Abstract Expression Atlas is EMBL-EBI’s resource for gene and protein expression. It sources and compiles data on the abundance and localisation of RNA and proteins in various biological systems and contexts and provides open access to this data for the research community. With the increased availability of single cell RNA-Seq datasets in the public archives, we have now extended Expression Atlas with a new added-value service to display gene expression in single cells. Single Cell Expression Atlas was launched in 2018 and currently includes 123 single cell RNA-Seq studies from 12 species. The website can be searched by genes within or across species to reveal experiments, tissues and cell types where this gene is expressed or under which conditions it is a marker gene. Within each study, cells can be visualized using a pre-calculated t-SNE plot and can be coloured by different features or by cell clusters based on gene expression. Within each experiment, there are links to downloadable files, such as RNA quantification matrices, clustering results, reports on protocols and associated metadata, such as assigned cell types. INTRODUCTION Expression Atlas (https://www.ebi.ac.uk/gxa/home) is an added-value online bioinformatics resource that includes a database, user interface and web-service enabling easy access to information about gene expression across species, tissues, cells, experimental conditions and diseases. It was originally developed in 2009 (1) as a resource to uniformly analyse, annotate and display the results of microarray and RNA-Seq experiments from the ArrayExpress archive (2). Since then, Expression Atlas has grown to process datasets from a variety of other sources and archives, such as NCBI’s Gene Expression Omnibus (GEO) (3), the European Nucleotide Archive (4) as well as controlled access datasets, such as GTEx. In addition, Expression Atlas serves as a long term portal for transcriptomics data generated by large-scale genomics studies such as Gramene (5), the comparative resource for plants (http://www.gramene.org/) and the Pancancer Analysis of Whole Genomes (PCAWG, https://docs.icgc.org/pcawg/). Expression Atlas has also integrated protein expression information via mass spectrometry proteomics datasets coming from the PRIDE database (6), and one dataset from the Human Protein Atlas. Over the last five years, processed and curated gene expression datasets have been regularly contributed to projects such as Open Targets (7), whilst Expression Atlas’ heatmap widget has been adopted by eight resources to display gene expression across different tissues within their websites (e.g. Reactome (8)). With recent advances in technologies for tissue dissociation and high-throughput sequencing at the single cell level, single cell RNA-Seq datasets are being generated and deposited into the public domain at an increasing pace. Since the last Expression Atlas update in 2017 (9), the main developments within Expression Atlas have therefore focused on the annotation, analysis, storage and display of single cell RNA-Seq studies. The Expression Atlas, previously consisting of a ‘baseline expression’ and a ‘differential expression’ component for expression at the level of tissues, has now acquired a third, ‘single cell’ component. This new component, the Single Cell Expression Atlas, was first launched in 2018 and currently includes 123 studies, including almost a million assays from 12 different species. Single Cell Expression Atlas employs analysis pipelines that process datasets from SMART-like and droplet-based experimental protocols in a standardised way and displays gene expression in different cells and cell types. Single Cell Expression Atlas is unique in that it provides to the life sciences community uniformly analysed and annotated single cell RNA-Seq data across multiple species. It enables easy access to meaningful results and standardised data files (see Figure 1), therefore providing easy input to software for further analysis gene of expression matrices, as well as supporting the development of computational methods for downstream analyses, such as cell type annotation. The field of single cell genomics is undergoing rapid developments in the range of experimental protocols. These result in large increases in data volumes, stressing the underlying infrastructure and causing frequent rounds of re-engineering. For example, bulk RNA-Seq experiments in Expression Atlas consist of just above hundred thousand assays across 809 studies, whereas 123 single cell studies comprise almost a million assays. Moreover, new analysis methods are constantly being generated and significantly improved, resulting in re-evaluations and updates of pipelines. Therefore, Single Cell Expression Atlas was launched to reside as a service within Expression Atlas with the expectation to be fully integrated and queried via Expression Atlas’ main gene and condition search in the near future, as the experimental protocols and analysis methods mature. [Figure: Screenshot of the ‘Browse Experiments’ page. A user can search through the experiments by species or other keywords, quickly link to their results pages or download raw, normalized data and metadata bundles from one or more studies.] SINGLE CELL EXPRESSION ATLAS Data Since its original launch in May 2018, the Single Cell Expression Atlas service has processed 123 studies consisting of 955,000 assays, from 12 different species, including Homo sapiens and model organisms such as Mus musculus, Arabidopsis thaliana and Drosophila melanogaster. All datasets are sourced from public archives, such as ArrayExpress and GEO. Metadata are curated in-house through a semi-automatic process of identifying the experimental factors, such as cell types, diseases or perturbations, followed by annotation with Experimental Factor Ontology terms (EFO) to describe the experimental comparisons for further processing. Samples are annotated with cell types, where these have been submitted or made available by the primary authors of these datasets. Cell types are then linked with terms from the Cell Ontology. Data analysis Analysis pipelines for Single Cell Expression Atlas comprise two parts: gene expression quantification and downstream analysis. Whilst quantification is accomplished using technology-dependent pipelines, downstream analysis is currently performed identically for all studies. Quantification of gene expression For SMART-like technologies (10) where data are available in cell-demultiplexed form, a bulk-like pipeline is used. This pipeline was originally implemented in iRAP (https://doi.org/10.1101/005991), but has recently been re-implemented using Nextflow (11). The pipeline comprises quality filtering, quality trimming and sequencing artifact removal with FASTX-Toolkit (http://hannonlab.cshl.edu/fastx_toolkit/), poly-A and uncalled base filtering with fastq-utils (https://github.com/nunofonseca/fastq_utils) and a contamination check. Paired-end reads are re-paired with fastq-pair (https://doi.org/10.1101/552885) after filtering. Filtered reads are then quantified with Kallisto (12), and cell-wise quantifications subsequently combined to a final matrix. For droplet technologies such as 10x (v2, v3) (13) and Drop-seq (14) with cell barcodes and unique molecular identifiers (UMIs), we employ Alevin, which is part of the Salmon package (15). Alevin handles barcode/UMI processing, quantification and the production of a quantification matrix. By default, Alevin filters cell barcodes based on a frequency distribution, but we found this difficult to operate in an unsupervised manner over many experiments. We therefore run Alevin without filtering and excluding only the lowest frequency barcodes, and subsequently apply the emptyDrops method of the DropletUtils package (16) to remove empty droplets. Where a study combines multiple libraries, the matrices resulting from each run are combined into a final matrix. This pipeline is under close review, with processes such as doublet removal being considered for inclusion in future iterations. Downstream analysis For all single cell