A multi-mapped read rescue strategy for peak-based gene regulatory analyses.

• Bioconda:

• PyPI:

git clone https://github.com/seqcode/allo.git
cd allo
pip install -e .

Using Allo requires a few pre-processing steps. In most ChIP pipelines, the default behavior of aligners is to assign multi-mapped reads to random locations within their mappings without retaining information on the other locations. Both Bowtie1/2 and BWA can be used for single-end. Unfortunately, BWA cannot be used for paired-end reads prior to Allo due to constraints in how it outputs multi-mapped reads. The following arguments should be used:

Bowtie1

#Single-end
bowtie -x INDEX -q FASTQ -S SAMOUT --best --strata -m 50 -k 50 -p THREADS
#Paired-end
bowtie -x INDEX -1 READ1 -2 READ2 -S SAMOUT --best --strata -m 50 -k 50 -p THREADS

Bowtie2

#Single-end
bowtie2 -x INDEX -q FASTQ -S SAMOUT -k 50 -p THREADS
#Paired-end
bowtie2 -x INDEX -1 READ1 -2 READ2 -S SAMOUT -k 50 --no-mixed --no-discordant -p THREADS

BWA

#BWA is not supported for paired-end processing due to constraints with multi-mapped read reporting
#Single-end
bwa mem -a INDEX FASTQ > SAMOUT

Finally, the output of the aligners must be sorted by read name in order to use Allo. If using Bowtie1 in paired-end mode, you may want to use samtools to filter unproperly mapped pairs. It is possible to use Allo with discordant or mixed pairs but Allo will treat them equally to proper pairs. Proceed accordingly.

samtools collate -o ALIGNEROUTPUT_SORT.SAM ALIGNEROUTPUT_FILTER.SAM

The basic command for Allo:

allo ALIGNEROUTPUT_SORT.SAM -seq PAIRED_OR_SINGLE -o OUTPUTNAME -m MIXED_OR_NARROW_PEAKS

Allo also accepts BAM files as input. See other options below..

During each run, Allo will create temporary files as it allocates the data. UM files are reads designated as uniquely mapped (has to be parsed in Bowtie2 or BWA). MM files are unallocated multi-mapped reads. AL files are allocated reads. Checking the size of the AL files during the run will give you an estimate of how many reads have already been allocated at that time.

Very short test files are supplied to make sure Allo runs to completion on your machine. Imports can take a minute so be patient. Using the paired-end example:

allo testRunPE.sam -seq pe

Allo adds a ZA tag to every MMR that is allocated. For reads that are allocated to regions that all contain 0 UMRs (random assignment), a ZZ tag is used instead. This allows users to remove reads that only map to zero UMR regions if they wish. The value within either tag corresponds to the number of places a read/pair mapped to. In order to get only uniquely mapped reads, grep could be used with the -v option to exclude lines with ZA or ZZ tags. On the same note, awk can used to filter reads with a specific number of mapping locations (can also be done with the -max option within Allo). Outside of adding these tags, Allo does not change anything within the read alignment columns for allocated reads.

Tip: It is recommended to run Allo on both the control and target sequencing files in order to balance out background in the samples. We recommend running Allo using the --random argument on the control file. This generally results in higher confidence peaks.

Please contact Alexis Morrissey ([email protected]) with any questions concerning Allo or open an issue here on GitHub.