Scater: pre-processing, quality control, normalization and visualization of single-cell RNA-seq data in R.

McCarthy DJ, Campbell KR, Lun AT, Wills QF. · 2017

Read Scater: pre-processing, quality control, normalization and visualization of single-cell RNA-seq data in R. by McCarthy DJ, Campbell KR, Lun AT, Wills QF. (2017) 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-pmc5408845.txt.

Introduction

Davis J. McCarthy, Kieran R. Campbell, Aaron T. L. Lun, and Quin F. Wills present a software package called scater in a 2017 paper in Bioinformatics. Single-cell RNA reading measures gene activity cell by cell, exposing rare cell types and fine steps in growth. Raw output is messy, though: bias, artifacts, and stray variation demand long cleaning before real study can start, and each lab handles that stage its own way.

Scater gives that stage one shared, careful path inside the R language. It cleans raw reads, checks quality, levels the samples, and plots the data in many forms suited to single cells. Its data format plays well with other tools and can serve as a base for future software. The package is free through Bioconductor, with install help and worked case studies.

Catalog record

CREATORMcCarthy DJ, Campbell KR, Lun AT, Wills QF.
YEAR2017
COLLECTIONPAPERS
FORMATTXT
FILE SIZE46 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC5408845

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Scater: pre-processing, quality control, normalization and visualization of single-cell RNA-seq data in R By Davis J McCarthy, Kieran R Campbell, Aaron T L Lun, Quin F Wills, Ivo Hofacker Published in: Bioinformatics DOI: 10.1093/bioinformatics/btw777 ABSTRACT Abstract Motivation Single-cell RNA sequencing (scRNA-seq) is increasingly used to study gene expression at the level of individual cells. However, preparing raw sequence data for further analysis is not a straightforward process. Biases, artifacts and other sources of unwanted variation are present in the data, requiring substantial time and effort to be spent on pre-processing, quality control (QC) and normalization. Results We have developed the R/Bioconductor package scater to facilitate rigorous pre-processing, quality control, normalization and visualization of scRNA-seq data. The package provides a convenient, flexible workflow to process raw sequencing reads into a high-quality expression dataset ready for downstream analysis. scater provides a rich suite of plotting tools for single-cell data and a flexible data structure that is compatible with existing tools and can be used as infrastructure for future software development. Availability and Implementation The open-source code, along with installation instructions, vignettes and case studies, is available through Bioconductor at http://bioconductor.org/packages/scater. Supplementary information Supplementary data are available at Bioinformatics online. 1 INTRODUCTION Single-cell RNA sequencing (scRNA-seq) describes a broad class of techniques which profile the transcriptomes of individual cells. This provides insights into cellular processes at a resolution that cannot be matched by bulk RNA-seq experiments (Hebenstreit and Teichmann, 2011; Shalek et al., 2013). With scRNA-seq data, the contributions of different cell types to the expression profile of a heterogeneous population can be explicitly determined. Rare cell types can be interrogated and new cell subpopulations can be discovered. Graduated processes such as development and differentiation can also be studied in greater detail. However, this improvement in resolution comes at the cost of increased technical noise and biases. This means that pre-processing, quality control and normalization are critical to a rigorous analysis of scRNA-seq data. The increased complexity of the data across hundreds or thousands of cells also requires sophisticated visualization tools to assist interpretation of the results. Numerous statistical methods and software tools have been published for scRNA-seq data (Angerer et al., 2015; Finak et al., 2015; Guo et al., 2015; Juliá et al., 2015; Kharchenko et al., 2014; Trapnell et al., 2014). However, all of these assume that quality control and normalization have already been applied. Fewer methods are available in the literature to perform these basic steps in scRNA-seq data processing (Ilicic et al., 2016). This issue is exacerbated by the diversity of scRNA-seq datasets with respect to the experimental protocol and the biological context of the study, meaning that a single processing pipeline with fixed parameters is unlikely to be universally applicable. Rather, software tools are required that support an interactive approach to analysis. This allows parameters to be fine-tuned for the study at hand in response to any issues diagnosed during data exploration. The provided functionality should also process the data in a statistically rigorous manner and encourage reproducible bioinformatics analyses. One of the most widely used frameworks for interactive analysis is the R programming language, extended for biological data analysis through the Bioconductor project (Huber et al., 2015). While Bioconductor packages have been widely used for bulk RNA-seq data, the existing data structures (like the ExpressionSet class) are not sufficient for scRNA-seq data. This is because they do not support data types that are specific to single-cell studies, e.g. cell–cell distance matrices for clustering. For larger studies, this also includes data beyond expression profiles such as intensity values from fluorescence-activated cell sorting, cell imaging data and information from epigenetic and targeted genotyping assays. Existing methods for processing and applying quality control to scRNA-seq data are similarly inadequate. In particular, current visualization methods designed for exploratory data analysis of bulk transcriptomic experiments are unsuited to datasets containing hundreds or thousands of cells. The large size of each dataset also favours methods such as kallisto (Bray et al., 2016) and Salmon (Patro et al., 2016) for rapidly quantifying gene expression, but support for the output of these methods is currently limited. Extensions to the current computational infrastructure are required to provide appropriate data structures and methods that can accommodate these rich scRNA-seq datasets for integrative analyses of expression and other assay data along with the accompanying metadata. Here we present scater, an open-source R/Bioconductor software package that implements a convenient data structure for representing scRNA-seq data and contains functions for pre-processing, quality control, normalization and visualization. The package provides wrapper functions for running kallisto and Salmon on raw read data and converting their output into gene-level expression values, methods for computing and visualizing quality-control metrics for cells and genes, and methods for normalization and correction of uninteresting covariates. This is done in a single software environment which enables seamless integration with a large number of existing tools for scRNA-seq data analysis in R. The scater package provides basic infrastructure upon which customized scRNA-seq analyses can be constructed, and we anticipate the package to be useful across the whole spectrum of users, from experimentalists to computational scientists. 2 METHODS, DATA AND IMPLEMENTATION 2.1 Case study with scRNA-seq data The results presented in the main paper and supplementary case study use an unpublished single-cell RNA-seq dataset consisting of 73 cells from two lymphoblast cell lines of two unrelated individuals. Cells were captured, lysed and cDNA generated using the popular C1 platform from Fluidigm, Inc. (https://www.fluidigm.com/products/c1-system). The processing of the two cell lines was replicated across two machines, with the nuclei of the two cell lines stained with different dyes before mixing on each machine. Cells were imaged before lysis, with an example image provided together with these data (see Case Study in Supplementary Material). Samples were sequenced with paired-end sequencing using the HiSeq 2500 Sequencing system (Illumina). RNA-seq reads were mapped to a custom genome reference, consisting of Homo sapiens GRCh37 (primary assembly from ftp://ftp.ensembl.org/pub/release-75/fasta/homo_sapiens/dna/, last accessed 14.08.2015), Epstein-Barr Virus type 1 (B95-8 strain, Accession NC_007605.1) and ERCC RNA spike-ins (ThermoFisher). Reads in fastq format were aligned with TopHat2 v2.0.12 (Kim et al., 2013), using Bowtie2 v2.2.3.0 (Langmead and Salzberg, 2012) as the core mapping engine (–mate-inner-dist 190 –mate-std-dev 40 –report-secondary-alignments) and other default parameters. Potential PCR duplicates were marked with Picard MarkDuplicates v1.92(1464). Reads mapping uniquely to annotated exon features were counted using htseq-count implemented in HTSeq, version 0.6.1p1 (Anders et al., 2015). Further case studies using scater on published data, for example from 3000 mouse cortex cells (Zeisel et al., 2015) and 1200 cells from early-development mouse embryos (Scialdone et al., 2016) are available at http://dx.doi.org/10.5281/zenodo.59897. All materials required to reproduce the results presented in this paper are available at http://dx.doi.org/10.5281/zenodo.60139. 2.2 Implementation The scater package is an open-source R package available through Bioconductor. Key aspects of the code are written in C ++ to minimize computational time and memory use, and the package scales well to large datasets. For example, consider the Macosko et al. (2015) dataset, which contains more than 44 000 cells. The core scater functions to create an SCESet object and calculate QC metrics took approximately two minutes to complete on an early 2015 MacBook Pro laptop with 2.9 GHz Intel Core i5 processor and 16

Read the complete text