Comparative Analysis of Single-Cell RNA Sequencing Methods

Abstract

1 Single-cell mRNA sequencing (scRNA-seq) allows to profile heterogeneous cell 2 populations, offering exciting possibilities to tackle a variety of biological and medical 3 questions. A range of methods has been recently developed, making it necessary to 4 systematically compare their sensitivity, accuracy, precision and cost-efficiency. 5Here, we have generated and analyzed scRNA-seq data from 479 mouse ES cells and 6 spike-in controls that were prepared with four different methods in two independent 7 replicates each. We compare their sensitivity by the number of detected genes and by 8 the efficiency with which they capture spiked-in mRNAs, their accuracy by correlating 9 spiked-in mRNA concentrations with estimated expression levels, their precision by 10 power simulations and variance decomposition and their efficiency by their costs to 11 reach a given amount of power. While accuracy is similar for all methods, we find that 12Smart-seq on a microfluidic platform is the most sensitive method, CEL-seq is the 13 most precise method and SCRB-seq and Drop-seq are the most efficient methods. 14 Our analysis provides a solid basis to choose among four available scRNA-seq 15methods and to benchmark future method development. 16 17 18 transcription allowing to process hundreds or thousands of scRNA-seq libraries in one 36 reaction, increasing the throughput of scRNA-seq library generation by one to three orders 37 of magnitude 9,13-15 . However, a systematic comparison of sensitivity, accuracy and precision 38 among such recently developed methods has not been performed yet. To this end, we have 39 generated and analysed 479 scRNA-seq libraries from mouse embryonic stem (mES) cells 40 cultured in two-inhibitor (2i/LIF) medium using four different methods run in two replicates 41 each (Fig. 1). 42 43 Results 44Generation and processing of 479 scRNA-seq libraries 45We have used the Smart-seq protocol on the C1 platform from Fluidigm (Smart-seq/C1) 46 that uses microfluidic chips to automatically separate up to 96 cells 7 . After lysis and the 47

Gene length and detection bias in single cell RNA sequencing protocols

“…To ensure the gene length bias is not due to the specific biology of the different cell types, we analysed four different mouse embryonic stem cell datasets generated using both UMI and full-length transcript protocols ( Buettner et al , 2015; Grün et al , 2014; Kolodziejczyk et al , 2015; Ziegenhain et al , 2016). When we combined all four datasets together (see methods) and performed principal components analysis, we noted that the cells clustered by dataset, with the UMI datasets on the left and full-length datasets on the right of the plot ( Figure 3a).…”

“…Four different mouse embryonic stem cell datasets were combined, two full-length transcript ( Buettner et al , 2015; Kolodziejczyk et al , 2015) and two UMI datasets ( Grün et al , 2014; Ziegenhain et al , 2016). ( a ) Principal component analysis plot (coloured by dataset) shows the major source of variation between the cells is the dataset, with the UMI datasets on the left and the full-length datasets on the right.…”

“…Mouse embryonic stem cells, Ziegenhain et al , 2016, UMI. We downloaded the molecule counts from GEO under accession GSE75790.…”

The Unmixing Problem: A Guide to Applying Single-Cell RNA Sequencing to Bone

“…Furthermore, while all current scRNA‐seq methods are limited to largely capturing only a fraction of highly expressed mRNAs from each cell, Drop‐seq captures fewer transcripts per cell than Smart‐seq, which can exacerbate analytic challenges created by sparse transcriptome sampling. Detailed analysis of the relative strengths of specific methods is available in recent methodologic comparison studies …”

Don't judge a cell by its cover: heterogeneity within early lymphoid progenitors

“…Several different methods have been developed, and Alberti‐Servera et al () chose to utilize the Fluidigm C1 platform and SMARTer chemistry for efficient single‐cell capture and library preparation. In combination with high sequencing depth, this protocol supports the detection of low expressed genes and transcription factors at a high enough coverage to allow for the separation of functionally very similar populations (Ziegenhain et al , ). A disadvantage is the large number of cells needed for each run, which makes the method difficult to use for rare cell populations.…”