Wrapper scripts and container are not portable

[moskalenko]

Hi,

I'm looking at CN_Learn to see if it can be installed on a compute cluster. However, I am concerned that the included wrapper scripts all have to be manually patched since they are not written to be portable. For example, even if though ${DOCKER_COMMAND} can be overridden in the user's environment to run a singularity container created from the provided docker container virtually all calls to executables in the container take the form of ${DIRECTORY_NAME}program whereas the convention is that all executables would be in standard path or paths specified in the docker container's environment, so all wrappers don't have to be painstakingly manually patched on every installation and upgrade.

At the same time the /opt/tools/CN_Learn/config.params in the container has hardcoded paths to directories likely present on the system(s) the author(s) work on like

REF_GENOME='/data/bx_references/GRCh37/human_g1k_v37.fasta'
DATA_BPCOV_DIR='/data2/RF_Model_SVIP/data/extract/bp_level_coverage/'

Regards,

Alex