4 Notes
1.scRNAseq method: An array of different scRNAseq approaches
are available to date: C1 Single Cell AutoPrep System (Flui-
digm), CEL-seq2 [12], Smart-seq2 [13], Drop-seq [14], and
InDrops [15], to name just a few. In these methods, single cells
are first captured using either droplet- or laminar flow based
microfuidic devices or fluorescence-activated cell sorting (FACS)
for the distribution of cells into reaction wells. The capture and
amplification of single cell RNA relies on the recognition of the
poly-A tail of mRNA molecules by reverse transcription primer.
Reverse transcribed cDNA molecules are subsequently amplified
by PCR (Smart-seq, Smart-seq2, Drop-seq) or in vitro transcrip-
tion (CEL-seq2, InDrops).
2.Variations in cell size and transcriptional activity: As men-
tioned, several scRNAseq methods are available to date. Some
of these methods are limited to a particular cell size range,
which sets limits to the coverage of the whole cell population,
as well as increases technical noise and complicates the normal-
ization procedures. It is known that larger or transcriptionally
more active cells with more mRNA molecules yield more com-
plex RNAseq libraries, due to technical reasons.
3.Complexity of the cell population: The approach described
here relies on the assumption that the data are a diverse collec-
tion of cells from several different cell types. Thus, while our
approach is ideal for the identification of different cell types
from a relatively balanced population, detection of rare and
underrepresented cell types is more challenging.
4.Critical parameters for building a gene expression atlas:
The two most critical parameters to consider when building a
gene expression atlas are the imaging resolution (pixel size) and
the minimum number of samples required to get a good aver-
aged expression for each individual marker (e.g., gene), as both
impact the final atlas resolution. A proper analysis of these and
other technical parameters is advised before acquiring the data
to build the atlas. For an example on one method to do this,see
Vergara et al. 2016 [5].
5.Cell preparation and cell viability: Good quality of scRNAseq
data is essential for reliable results. Low complexity scRNAseq
samples map to spatial reference with poor confidence and
broadly. By our experience, good cell viability (>90% viable
cells) and short preprocessing time of single-cell preparations is
key to good quality sequencing samples.
We encourage testing different treatments to obtain single
cell suspensions. The cell dissociation procedure should be
optimized for speed and completeness, while minimizing the122 Kaia Achim et al.