Profiling

The speed of execution of isoslam process can be investigated using Profiling which analyses which functions and methods take up the most processing time. In order to undertake profiling the cProfile standard library can be used. Visualisation of the results can be aided using the SnakeViz package.

Performing Profiling

You need a set of sample files to process in order to undertake profiling. Here we use the sample files that are included as part of the test suite used in development that can be found under tests/resources. If you do not yet have these locally you should git clone the repository and install it in a clean virtual environment with the development dependencies.

git clone git@github.com:sudlab/IsoSLAM.git
cd IsoSLAM
pip install -e .[dev]

We now make a tmp/test-YYYYMMDD directory and copy the necessary files here.

mkdir -p tmp/test-$(date +%Y%m%d)                            # Uses the current date
cp -r tests/resources/{bam,gtf,bed,vcf} tmp/test-20250221    # Modify to reflect the current date

We can run profiling on these samples using the following which writes the profiling to isoslam-YYYYMMDD.prof.

cd tmp/test-20250221
python -m cProfile -o isoslam-$(date +%Y%m%d).prof $(isoslam process \
   --bam-file bam/sorted_assigned/d0_no4sU_filtered_remapped_sorted.sorted.assigned.bam \
   --gtf-file gtf/test_wash1.gtf \
   --bed-file bed/test_coding_introns.bed \
   --vcf vcf/d0.vcf.gz)

You can verify output has been produced using parquet-tools which is part of the dev dependencies.

parquet-tools show output/results.parquet

The profiling data should have been written to isoslam.prof.

head isoslam.prof

Visualisation of Profiling

To visualise the results of profiling you can invoke snakeviz with the .prof file that has been generated.

snakeviz isoslam-<YYYYMMDD>.prof

This should launch a new browser tab with the icicle where the amount of time spent within a function is proportional to the size of the bar.