higher throughput, and more reproducibility. In 2015, [22, 23]
introduced droplet-based methods where single cells are placed in
droplets using microfluidics and beads with UMIs to uniquely label
RNA molecules in each cell.
Currently a number of platforms are available to choose
between, each with its advantages and disadvantages. Platforms
differ from each other by either method of RNA quantification,
or by method of cell capture. RNA expression is quantified by
measure of either full length cDNAs or by tag-based UMIs.
There are three methods of cell capture, microwell-based, microfluidic-based, and droplet-based. With the various options, it may
seem difficult to determine which method is best, and the answer is
it depends on the question being asked. Ziegenhain et al. [24] and
Svensson et al. [25] realized this and so to assist you in making an
informed decision they compare and contrast the common scRNAseq techniques’ accuracy, sensitivity, precision, power, and cost
efficiency. Based on their findings, Smart-seq2, had the best sensitivity, accuracy, precision, and the lowest gene dropout rate, however this approach provides relatively low throughput compared to
droplet-based methods that are not as sensitive but significantly less
costly. Smart-seq2 currently is the best option for increased
sequencing depth but for a smaller number of cells, as cost can be
quite considerable. If willing to sacrifice some sequencing depth,
drop-seq is the most cost efficient of the methods, but requires a
tedious multi day protocol to be performed. Labs and sequencing
centers also are adapting commercial platforms that include Fluidigm’s C1 microfluidic chip, Wafergen ICELL8, BioRad’s ddSEQ,
and perhaps the most popular, 10Â Genomics Chromium. Other
alternatives utilize combinatorial indexing such as sciRNA-seq,
while SPLiT-seq utilizes a split and pool method of barcoding
cells within wells [4, 26]. These allow for higher throughput and
cost efficiency than 10Â and Drop-seq, however, the sample preparation takes longer, and there is a potential for introduction of
sampling bias. In addition, the cell quality reportedly is a bit lower
than 10Â and Drop-seq. With all these options, it can be difficult to
identify which method is best, for your research question. For a
process such as EMT which has a temporal component, and for a
process that occurs within an in vivo model (in our case, the sea
urchin), we sought a method that could process many single cells
with the best depth possible. To satisfy such a requirement, 10Â
Genomics was our choice of platform. Following library construction of single cells via 10Â Genomics protocol, cells are sequenced
at ~50 k reads per cell and using a 150 bp paired end Illumina run.
Similarly, other single cell library preparation protocols utilize Illumina’s paired end sequencing but may have different run length of
75, 125, 250 bp and more. Depending on the number of cells and
the run length, a variety of options will be available using Illumina.
For example, using a total of 1 billion PE reads on the NovaSeq
308
Abdull J. Massri et al.
introduced droplet-based methods where single cells are placed in
droplets using microfluidics and beads with UMIs to uniquely label
RNA molecules in each cell.
Currently a number of platforms are available to choose
between, each with its advantages and disadvantages. Platforms
differ from each other by either method of RNA quantification,
or by method of cell capture. RNA expression is quantified by
measure of either full length cDNAs or by tag-based UMIs.
There are three methods of cell capture, microwell-based, microfluidic-based, and droplet-based. With the various options, it may
seem difficult to determine which method is best, and the answer is
it depends on the question being asked. Ziegenhain et al. [24] and
Svensson et al. [25] realized this and so to assist you in making an
informed decision they compare and contrast the common scRNAseq techniques’ accuracy, sensitivity, precision, power, and cost
efficiency. Based on their findings, Smart-seq2, had the best sensitivity, accuracy, precision, and the lowest gene dropout rate, however this approach provides relatively low throughput compared to
droplet-based methods that are not as sensitive but significantly less
costly. Smart-seq2 currently is the best option for increased
sequencing depth but for a smaller number of cells, as cost can be
quite considerable. If willing to sacrifice some sequencing depth,
drop-seq is the most cost efficient of the methods, but requires a
tedious multi day protocol to be performed. Labs and sequencing
centers also are adapting commercial platforms that include Fluidigm’s C1 microfluidic chip, Wafergen ICELL8, BioRad’s ddSEQ,
and perhaps the most popular, 10Â Genomics Chromium. Other
alternatives utilize combinatorial indexing such as sciRNA-seq,
while SPLiT-seq utilizes a split and pool method of barcoding
cells within wells [4, 26]. These allow for higher throughput and
cost efficiency than 10Â and Drop-seq, however, the sample preparation takes longer, and there is a potential for introduction of
sampling bias. In addition, the cell quality reportedly is a bit lower
than 10Â and Drop-seq. With all these options, it can be difficult to
identify which method is best, for your research question. For a
process such as EMT which has a temporal component, and for a
process that occurs within an in vivo model (in our case, the sea
urchin), we sought a method that could process many single cells
with the best depth possible. To satisfy such a requirement, 10Â
Genomics was our choice of platform. Following library construction of single cells via 10Â Genomics protocol, cells are sequenced
at ~50 k reads per cell and using a 150 bp paired end Illumina run.
Similarly, other single cell library preparation protocols utilize Illumina’s paired end sequencing but may have different run length of
75, 125, 250 bp and more. Depending on the number of cells and
the run length, a variety of options will be available using Illumina.
For example, using a total of 1 billion PE reads on the NovaSeq
308
Abdull J. Massri et al.