Pyomo-supported ILP solver

[Riyue Bao]

Hello,

OptiType returned the following message while calling the module model.py:

Invalid option '-s'; try /group/bioinformatics/software/GLPK/4.61/bin/glpsol --help
ERROR: "[base]/site-packages/pyomo/opt/base/solvers.py", 599, solve
	Solver (asl) returned non-zero return code (1)
ERROR: "[base]/site-packages/pyomo/opt/base/solvers.py", 602, solve
	See the solver log above for diagnostic information.

This is my command:

module load anaconda3
source activate python2.7

module load GLPK/4.61 HDF5/1.10.0-patch1 samtools/1.2 bwa/0.7.15 sambamba/0.5.6

python ./OptiTypePipeline.py -i ./test/exome/NA11995_SRR766010_1_fished.fastq ./test/exome/NA11995_SRR766010_2_fished.fastq -o test -d -v

source deactivate

I was not able to install RazerS and CPLEX on our server, hence went for bwa and GLPK instead. I modified OptiTypePipeline.py a little bit to take in bwa alignment commands instead of the default RazerS .

The command-line solver for GLPK is glpsol, which I also changed in config.ini. Could that be the reason why the error was reported? The modified OptiTypePipeline.py, config, commands and log files are attached here if helpful myfiles.zip

Any help would be highly appreciated. Thanks much in advance :)

Best,
Riyue (Sunny)
The University of Chicago

[b-schubert]

Hi Riyue,

Could you change the solver name to glpkand try it again. Also make sure that glpk and bwa can be found in PATH.

I'm also not sure that bwa is a good alternative as OptiType needs an all-bestmapping approach.

If you have problems with installing the requirements by hand, you could try out our Docker image which comes with a pre-configured RazerS3 and a ILP solver (see installation instruction for more infos).

[Riyue Bao]

Hi Benjamin,

Thanks for the quick response! I created a symbolic link glpk -> glpsol and it worked. Yay!!

Here is the result from running the test exome fastq files and the coverage plot
2017_04_17_13_27_05_coverage_plot.pdf

	A1	A2	B1	B2	C1	C2	Reads	Objective
0	A*24:03	A*24:03	B*08:12	B*57:03			31.0	29.844

I agree bwa may not be a good alternative for razers3. Docker would be an excellent option but unfortunately our server does not have it installed. Would you have any recommendations on other aligners for OptiType, e. g. bowtie2 or novoalign? bowtie2 has the alignment mode -a mode: search for and report all alignments and novoalign has -r all. Would these be potential good alternatives? Thanks much in advance!

Best,
Riyue

[b-schubert]

Uh, the result is definitely incorrect. This is probably due to bwa. You could try Yara (page currently down - https://www.seqan.de/yara/), which is the successor of RazerS3 and should be natively supported by OptiType.

What exactly went wrong during RazerS3's installation?

[Riyue Bao]

Thanks Benjamin. Yup that was due to the default alignment settings of bwa, which does not report all best alignments and filter out multitargets, and is not suitable for mapping to HLA loci.

The issue with RazerS3's installation was the gcc version - but our system adm helped fix it yesterday!! :) I compiled yara and razers3 and they both worked (BTW if anyone is interested in compiling individual tools from seqan, the installation instructions on yara github is super helpful).

This is the new result generated with aligner razers3:
dna fastq files & -d: 2017_04_19_12_50_43_coverage_plot.pdf

	A1	A2	B1	B2	C1	C2	Reads	Objective
0	A*01:01	A*01:01	B*08:01	B*57:01	C*07:01	C*06:02	1156.0	1135.192

rna fastq files & -r: 2017_04_19_12_53_22_coverage_plot.pdf

	A1	A2	B1	B2	C1	C2	Reads	Objective
0	A*31:01	A*68:01	B*40:01	B*51:01	C*15:02	C*03:04	132.0	128.436

The results make a lot more sense now ... :)

[Riyue Bao]

Hi just want to follow up that I have run all my samples by OptiType and the results look great. For each exome it took 5 to 15 mins to run with 8 threads from raw fastq to final output - super fast!