Simplify run_bwa_on_sample

[averagehat]

Is this still necessary? I know we don't need to filter on platforms any more. Not sure if compiling reads is necessary?

[necrolyte2]

I believe this handles the Paired and Unpaired reads though which is
necessary since if a 454 or ion torrent run is done without paired end you
will need to do 2 separate mappings and join the bams together after

On Thu, Jul 30, 2015 at 4:26 PM, Mike Panciera [email protected]
wrote:

Is this

ngs_mapper/ngs_mapper/run_bwa.py

Lines 29 to 66 in eba2386

	preads = reads_by_plat( args.reads )
	logger.debug( "Reads parsed by platform: {0}".format(preads) )
	reads = []
	for plat in args.platforms:
	if plat in preads:
	reads += preads[plat]
	# Creates reads/F.fq, reads/R.fq, reads/NP.fq
	readdir = join(tdir,'reads')
	os.makedirs( readdir )
	reads = compile_reads( reads, readdir )
	if not reads:
	raise Exception( "Somehow no reads were compiled" )
	if os.path.isdir( args.reference ):
	cwd = os.getcwd()
	os.chdir( tdir )
	ref = join( tdir, compile_refs( args.reference ) )
	os.chdir( cwd )
	else:
	ref = args.reference
	# Keeps track so we know to merge bams later if it is 3
	merge = 0
	if reads['F'] is not None:
	merge += 1
	pairedsai = bwa_mem( reads['F'], reads['R'], ref, join(tdir, 'paired.sai'), t=args.threads )
	if isinstance(pairedsai,int):
	raise BWAError("There was an error running bwa")
	pairedbam = ngs_mapper.bam.sortbam( ngs_mapper.bam.samtobam( pairedsai, PIPE ), join(tdir, 'paired.bam') )
	#bam.indexbam( pairedbam )
	if reads['NP'] is not None:
	merge += 2
	nonpairedsai = bwa_mem( reads['NP'], ref=ref, output=join(tdir, 'nonpaired.sai'), t=args.threads )
	if isinstance(nonpairedsai,int):
	raise BWAError("There was an error running bwa")
	nonpairedbam = ngs_mapper.bam.sortbam( ngs_mapper.bam.samtobam( nonpairedsai, PIPE ), join(tdir, 'nonpaired.bam') )

still necessary? I know we don't need to filter on platforms any more. Not
sure if compiling reads is necessary?

—
Reply to this email directly or view it on GitHub
#156.

[averagehat]

I am working on simplifying the pipeline up to the basecaller + runsample.py. That should make things smaller and easier to maintain and change.

Probably I may also want to change tests like test_project_directories_have_expected_files and test_consensus_mutations so that they do not take so long to run.

I want to use some declarative framework like Make or luigi. The only branching has to do with paired/unpaired and quality encoding, which can be determined statically (i.e. before the pipeline actually runs), so it should work out well.

If all goes well it will be easy to add steps like BAM cleaning before the base-calling or plotting happens as long as those two parts are connected reasonably.

[averagehat]

Closing in favor of #155, #164