using python to write bioinformatics pipelines tutorial

Question

I was wondering if there is a tutorial or a small code snippet to understand how to write bioinformatics pipeline using python, for example

1. use a aligner (say hisat)
2. get the output and process it using samtools

I was able to use subprocess from python2.7 for this purpose using samtools but i am not able to link both the processes.i.e given path(which I can use argparse) for directory with fastq files output would be processed bam.

sample code for samtools sam to bam :

import subprocess
subprocess.run(['samtools','view', '-bS',
                '../some.sam',
                '>',
                '../some.bam'])

Answer 1

I agree that using a specialized tool is probably a good idea.

Nevertheless if you want to stick with Python, I suggest using plumbum instead of subprocess. It has a very nice syntax for this kind of problems.

from plumbum import local

# load a command
samtools = local['samtools']

# you can then easily redirect
(samtools['view', '-bS', '../some.sam'] > '../some.bam')()

# it is also easy to pipe 
hisat2 = local['hisat2']
chain = hisat2['-x', 'genome.index', '-1', 'reads_R1.fq.gz', '-2', 'reads_R2.fq.gz'] | samtools['sort'] > 'reads_vs_genome.bam'

chain()

Answer 2

Taking a different tack from other answers, there's lots of tools for pipelines in Python. Note: there was a time when people would use "pipeline" to refer to a shell script. I'm talking about something more sophisticated that helps you decompose an analysis into parts and runs it robustly.

• Snakemake is my favourite. It's (nearly) pure Python and can generate reports.
• Nextflow is growing in popularity and is pretty straightforward
• Ruffus used to be reasonably popular, seemed fine when I used it
• Bpipe is for bioinformatics
• Airflow is a more industrial "big" solution but in wide use
• See this big list of pipeline systems

I'm sure you can find something there.

Answer 3

BioPython has some good tools for processing reads and alignments. http://biopython.org/DIST/docs/tutorial/Tutorial.html

There is a python library wrapping samtools so many of the samtools calls can be used directly as python objects and calls https://pysam.readthedocs.io/en/latest/

I would use subprocess to call the aligner and specify the output to a bam file that you have named and then read that bam file with pysam to do the analysis that you are interested in.

If you provide more specifics about the analysis or aligners I can help you more but you are in the right track just need to connect the two calls together.

example:

import subprocess
import pysam
import os 

for root, dirs, files in os.walk(rootPath):
    for filename in fnmatch.filter(files, ".fastq"):

        subprocess.run([`hisat`, `-f`, filename `-o`, `some.bam`])

        #get depth
        depth = pysam.depth('some.bam')

Answer 4

If you're just doing alignment and conversion to a sorted BAM file, there's no need to run it through python. A simple pipe on the unix command line works just as well (and probably runs faster):

hisat2 -x genome.index -1 reads_R1.fq.gz -2 reads_R2.fq.gz | 
  samtools sort > reads_vs_genome.bam

Answer 5

Have a look at luigi for writing pipelines in python. You can found BioLuigi if you like :-P