Adaptation to BD Rhapsody CB/UMI Structure

[Kai-12345]

Hi guys,
we are planning on running experiments using the BD Rhapsody scRNA-Seq platform on ONT. Would it be possible to adapt flexiplex for such approach. The main difference to the 10x barcode structure is the use of three separate barcode segments (CLS1, CLS2, CLS3), each from its own whitelist, separated by fixed linker sequences. Cell identity is the combination of all three. The "rules" which apply to barcode design are as follows
Each cell label section (CLS) has a whitelist of 384 possible 9 base sequences. All the capture oligos from a single bead will have the same cell label:

--CLS1---|-L1-|--CLS2---|-L2-|--CL3---|--UMI---|-CaptureSequence-

Further, each bead has two different oligo-structures (one for polyA-capture and another used in TCR/BCR sequencing having a TSO-capture sequence instead of the polyT):

polyT capture oligo:
[Enh_insert 0-3 bases] + [B384_cell_key1] + [Enh_linker1] + [B384_cell_key2] + [Enh_linker2] + [B384_cell_key3] + [8 random base UMI] + [25 base polyT capture]

5prime capture oligo:
[Enh_5p_primer] + [B384_cell_key1] + [Enh_5p_linker1] + [B384_cell_key2] + [Enh_5p_linker2] + [B384_cell_key3] + [8 random base UMI] + [Tso_capture_seq]

More information on how BD handles Barcode assignment can be found in the official pipeline documentation:
https://bd-rhapsody-public.s3.amazonaws.com/CellLabel/rhapsody_cell_label.py.txt

It would be great to have such tool available!

[nadiadavidson]

Hi,

We usually deal with multiple barcodes by running flexiplex multiple times (generally piping the input from one run to the next, so it's fast and no intermediate files are created) eg.:

cat raw_read.fastq | flexiplex -b "?????????" -k CLS1_list.txt -x [L1] -f 1 -e 1 -n BC1 | flexiplex -b "?????????" -k CLS1_list.txt -x [L2] -f 1 -e 1 -n BC2 | flexiplex -b "?????????" -k CLS3_list.txt -u "????????" -x [10bp of polyT or TSO capture seq] -f 1 -e 1 -n BC3 > demultiplexed_read.fastq

This would give you read IDs with structure:
@[BC3][UMI]#[BC2]#[BC1]_#[original read ID]

You may need to adjust some of the parameters like f (maximum edit distance of flanking sequence) and e (maximum edit distance to barcode), as well as how to anchor (the -x part) etc. There is a way to anchor to the start of reads as well if needed:

add_start() {
awk 'NR % 4 == 2 || NR % 4 == 0 { print "^" $0 } NR % 4 == 1 || NR % 4 == 3 { print }'
}
cat raw_read.fastq |
add_start | flexiplex -x "^" -b "?????????" -k CLS1_list.txt -x [L1] -f 0 -e 1 -n BC1 |
add_start | flexiplex -x "^" -b "?????????" -k CLS1_list.txt -x [L2] -f 0 -e 1 -n BC2 |
add_start | flexiplex -x "^" -b "?????????" -k CLS3_list.txt -u "????????" -x [10bp of polyT or TSO capture seq] -f 1 -e 1 -n BC3 > demultiplexed_read.fastq

This is obviously a little complicated and something we'd like to simplify in future releases. If you'd like some help optimising the command, and are happy to share a sample of the reads, we can suggest the exact parameters/command to use.

We've been asked similar questions by a few other people in the past, so you might also find it informative to check some of the closed github issues related to muti-barcodes as well.

All the best,
Nadia