Quick Start

pradosj/sindbis is a docker container implementing a pipeline to parse FASTQ reads sequenced from SINDBIS infected cells.

The pipeline takes 2 inputs:

1. a gzipped FASTQ file with extension <file>.fastq.gz. This file should contain 106bp-long-reads from SINDBIS infected cells.

2. a tab separated file <file>.index.tsv describing how to demultiplex sequences into the dissected areas. The first column must match the pattern pupName_dissectionName, and the dissectionName of the injection site must contains the string Inj.

Assuming the 2 input files exist, the pipeline can be run using the following command line into a terminal:

docker run --rm -v $(pwd):/export pradosj/sindbis <file>.sindbis/all

Example .index.tsv

Here is an example of input file for .index.tsv

name	sequence
Q_S1_Inj	ATCACG
Q_S1c	CGATGT
Q_M1	TTAGGC
Q_M1c	TGACCA
Q_S2	ACAGTG
Q_S2c	GCCAAT
Q_Str	CAGATC
Q_Strc ACTTGA

Installation

To run the pipeline, you need to install Docker, and make sure enough memory resources are allocated to Docker (mine is set to 12Gb). To check the memory allocation: launch docker and in the docker menu navigate to Preferences... >> Resources >> ADVANCED.

Parameters

You can set the desired filtering parameter using parameter FILTER=10. This parameter control the minimum amount of barcode required in the Injections sites (default 100).

docker run --rm -v $(pwd):/export pradosj/sindbis FILTER=10 <file>.sindbis/all

Pipeline output

The pipeline is implemented using Makefile rules, and generate multiple files in the ".sindbis/" subdirectory

1. demuxmap.tsv: a table mapping the demultiplexing index sequence to a MOUSE_DISSECTION name. The index is a 6bp nucleotide sequence that will be matched against the reads sequence at position 89-94 for demultiplexing.
2. *.fastq.gz: demultiplexed fastq files
3. demux_report.txt: summary statistics with the number of read in each demultiplexed region
4. dissections.txt: the list of dissection names extracted by parsing demux_report.txt
5. links.txt: pairs of Injection-Target dissections to link
6. *.umi.class.tsv.gz: a sorted table of UMI corrected barcodes sequence classified by type. Each row in the table give the number of time the barcode is seen with the same UMI sequence. The umi is the 12bp sequence extracted from the reads starting at position 85. bc32 is a 32bp barcode sequence found in the 32 first bp of the reads. class is a classification of the reads according to the content of bc32 sequence: if the last 8bp sequence of the barcode match with the sequence ATCAGTCA with a tolerance of 1 mismatch, the read is classified as spike, otherwise if the last 2bp are YY it is classified as viral.
7. *.umi.class.viral.fasta: the set of viral barcodes (bc32) found in *.umi.class.tsv.gz. The header of the fasta sequence also contains the umi-corrected count of barcode (i.e. the number of time the barcode is seen with distinct umi sequence).
8. MOS3P21_MO_Inj.umi.class.viral.self.bam: the sequence of above .fasta file mapped onto itself with bowtie (allowing for 2 mismatches).
9. *.bam.clusters.tsv: result of error correction on the barcode sequences.
10. *.bam.clusters.filter10.fasta: the error corrected list of barcodes found in the injection site, with at list 10 counts.
11. *.self.bam.clusters.filter10.bowtie_aln/: a directory with the result of mapping the barcodes in the targets dissections onto the barcodes found into the injection site. a. *.self.bam.clusters.filter10.bowtie_aln/*.bam: result of mapping the barcodes of a given target onto the barcode of the injection site (mapping done with bowtie, with a tolerance of 2 mismatches). b. *.self.bam.clusters.filter10.bowtie_aln/*.bam.tsv.gz: identical to ../*.umi.class.tsv.gz with the addition of 2 columns where barcodes as been mapped onto barcodes of the injection site.

Generate Barcode Count Matrix for R

For R/Bioconductor users, we have introduce a command to generate a SummarizedExperiment object containing umi-corrected barcode counts. This command looks recursively into the given directory for all files matching pattern *.umi.class.viral.bam.tsv.gz; read them to extract the number of barcode into each target that match a barcode of the injection site.

docker run --rm -v $(pwd):/export pradosj/sindbis <directory>/umi_corrected_count_matrix.rds