Fast searches of large collections of single cell data using scfind

Abstract

Single cell technologies have made it possible to profile millions of cells, but for these resources to be useful they must be easy to query and access. To facilitate interactive and intuitive access to single cell data we have developed scfind, a search engine for cell atlases. Using transcriptome data from mouse cell atlases we show how scfind can be used to evaluate marker genes, to perform in silico gating, and to identify both cell-type specific and housekeeping genes. Moreover, we have developed a subquery optimization routine to ensure that long and complex queries return meaningful results. To make scfind more user friendly and accessible, we use indices of PubMed abstracts and techniques from natural language processing to allow for arbitrary queries. Finally, we show how scfind can be used for multi-omics analyses by combining single-cell ATAC-seq data with transcriptome data. easy to interface with other widely used resources, e.g. the genome wide association study (GWAS) catalog (MacArthur et al., 2017) , the gene ontology (GO) knowledgebase (The Gene Ontology Consortium, 2017) , the disease ontology Medical Subject Headings (MeSH) (Sewell, 1964) , and collections of clinically important genetic variants (Cariaso and Lennon, 2012;Landrum et al., 2016) .Many of the challenges in working with single cell data stem from its large size. Typically, a single cell RNA-seq (scRNA-seq) dataset is represented as a gene-by-cell expression matrix with ~20,000 genes and may include millions of cells. For epigenetic data, e.g. scATAC-seq, the situation is often worse as the matrix can have many more rows, each representing a peak. Even though the matrix is sparse, working with such a dataset places high demands on computer hardware. Many operations are not only time consuming, but require advanced bioinformatics skills. To allow for both interactive and high-throughput queries of a large cell atlas, novel algorithms and data structures are required.To ensure that large single cell datasets can be accessed by a wide range of users, the underlying software must (i) allow for complex queries, (ii) take full advantage of the single-cell resolution of the data, (iii) support different types of assays besides RNA-seq, (iv) be fast and easy to use, (v) be possible to run interactively through a web browser to access large repositories of data, (vi) be possible to run locally by a user interested in analyzing their own data. However, none of the computational tools available today for collections of scRNA-seq datasets, e.g. Panglao (Franzén et al., 2019) , the UCSC Cell Browser (Haeussler et al., 2019) , scRNASeqDB (Cao et al., 2017) , SCPortalen (Abugessaisa et al., 2018) , and the EBI Single Cell Expression Atlas (Athar et al., 2019) , provide the required functionality and versatility.Here, we present scfind, a search engine that makes single cell data accessible to a wide range of users by enabling sophisticated queries for large datasets through an interface which is both very fast and familiar to users from any backg...

Integrated Transcriptome and Network Analysis Reveals Spatiotemporal Dynamics of Calvarial Suturogenesis

“…To aid cell type and cell function classifications we performed ranked gene enrichment analysis with the R extension of gProfileR (FDR ≤ 0.05) ( Reimand et al, 2007 ). In addition, we queried the top markers of each cell cluster against indexes built for the Mouse Cell Atlas ( Han et al, 2018 ) for bone marrow, brain, embryonic mesenchyme, embryonic stem cells, mesenchymal stem cells, and neonatal calvaria datasets using the search index Scfind ( Lee et al, 2019a ). We identified the hypodermis and dura mater clusters by reference to recent single cell studies of these tissues ( DeSisto et al, 2019 ; Driskell et al, 2013 ; Philippeos et al, 2018 ; Sennett et al, 2015 ).…”

Extensive remodeling of the extracellular matrix during aging contributes to age-dependent impairments of muscle stem cell functionality

“…To investigate which cell types contribute the most to the remodeling of the MuSC niche during aging, we integrated singlecell RNA-seq data from adult skeletal muscle (Giordani et al, 2019) with our proteome data. To identify the cell type(s) expressing the identified ECM proteins affected by aging, we used a hypergeometric test, as implemented in scfind (Lee et al, 2019; see STAR Methods for details). Thereby, we mapped the majority of ECM proteins changed in aging skeletal muscle to at least one cell type (53/67, 79%; Figure S6D).…”

“…The data from Giordani et al (2019) was downloaded from NCBI Gene Expression Omnibus (GEO) (Edgar et al, 2002) (accession GSE110878). Using the cell type labels assigned by the original authors, an scfind object was constructed (Lee et al, 2019). The selected 67 ECM proteins affected by aging were queried individually against the index and the enrichment as defined by a hypergeometric test, as displayed in Figure 3E.…”

Fast searches of large collections of single-cell data using scfind

“…The code for scfind is available at github.com/hemberg-lab/scfind and the code for generating the figures in this manuscript is available at https://github.com/hemberg-lab/scfind-paper-figures A Code Ocean capsule of the tool is provided ( https://doi.org/10.24433/CO.2453077.v1 ) 62 .…”