- Installing PeakSegFPOP is straightforward. I just followed the instructions and was able to get it installed.
$ Rscript test_demo.R
$ Rscript pipeline.R ./test/demo/
I executed the above commands and was able to view the results which were generated in ./test/demo directory. The results were the same results as shown on http://cbio.mines-paristech.fr/~thocking/PeakSegFPOP-input-data/
The generated results are also available in this repository.
Missing parts in PR:
- Conversion of labels
- Creating problems for each sample before computing targets
- Summarizing peak prediction
- Also, all the mentioned steps in the
stepmethod don't have definitions. So the PR is not complete in itself. Its just a skeleton code.
- Interdependent steps needs have tests run before the next step is executed to check that the previous step has returned output data or files necessary for the following step.
- If a particular step fails, that step has to be re run. (Having pre defined return codes would be helpful)
- When a particular step keeps failing, either the pipeline has to be aborted and re run, or completely aborted telling that it has failed.
- Steps which repeat, need to have time stamp based output naming convention, so that dependent steps would take the latest output data.
- Optional delays between jobs for better reliability.
- Jobs which are not dependent, if their duration and importance are known before hand, job scheduling algorithms can be applied for better reliability.