Scripts to reproduce the results from the GraphMap paper.
All tests on real data are implemented in a single script src/run-realdata.py.
Tests are implemented as individual functions that are called sequentially one after the other.
Simulations are implemented in our other repository aligneval which will be downloaded and setup automatically.
Installation
The tools that are required for testing can be set-up using the following command:
src/run-realdata.py setup-tools
This will automatically download and install: PyVCF, Tabix, VCFtools, LoFreq, samscripts, Mutatrix and Bamsurgeon.
Sudo is required to install the first three on the list.
The above command will also install aligneval which is our repo for evaluating alignments on simulated data.
Aligneval includes the wrapper scripts for each mapper/aligner required to run the tests. It will automatically download and install all aligners, but will not start generating simulated data right away. For this, the setup-simdata is used.
Manual setup required for GATK and PicardTools because of their download/licencing process. Please obtain GenomeAnalysisTK-3.4-46 and picard-tools-1.138 and place them in the tools folder.
Make sure that these are accessible: graphmap-reproduce/tools/picard-tools-1.138/picard.jar and graphmap-reproduce/tools/GenomeAnalysisTK-3.4-46/GenomeAnalysisTK.jar.
Running tests on simulations
src/run-simulated.sh
Everything should setup and run automatically.
This script will clone and install Aligneval, unpack the references, unpack the pre-simulated data, generate some simulated data which wasn't packed, and download the hg19 GRCh37 reference.
The entire process might take some time, especially if you consider the slowness of some mappers.
The user will be prompted several times during the installation of Aligneval, and twice at the very beginning of the alignment process. The prompts at the beginning of alignment are to test/skip the slower mappers.
Final results will be placed in the graphmap-reproduce/results folder.
Running tests on real data
This is a bit more trickier, because the real data is huge. Most of the data is provided as raw nanopore .fast5 files, and unpacking them might take over 800GB of disk space.
The entire process may take more than a day on a computer with a good Internet connection.
To initiate this process, run:
src/setup-realdata.py
Warning - at this point, automatic setup of reference sequence has yet to be implemented. One can download them manually from NCBI and place them in the correct paths (specified in src/run-realdata.py).
To run the alignment and the evaluation of the results, run:
src/run-realdata.py run
Please make sure that all reads and references are downloaded and correctly placed.
Sudo is required to run the tests, as the memory/time measurements require access to the root folders.