[RFC] Enabling Cython-based PDB parser backend for speed improvements

[a-r-j]

Describe the workflow you want to enable

Currently, the pure-python of PDB parsing in BioPandas is quite slow - certainly too slow for highthroughput structural bioinformatics or ML.

Describe your proposed solution

I have written a Cython-based implementation (CPDB) which is considerably faster and would like to set this as the default parsing backend. As it stands, I believe this to be one of the fastest (if not the fastest) available PDB parser for Python.

Performance comparison

However, given BioPandas' widespread usage, I am unclear if distributing this with a Cython component will lead to dependency problems for users.

Describe alternatives you've considered, if relevant

Speeding up the passage of time

Additional context

[rasbt]

@a-r-j This is super cool.

Btw. perhaps we don't need to worry about it extra dependencies here because NumPy already uses Cython (https://github.com/numpy/numpy/blob/main/build_requirements.txt), and pandas is build on NumPy, and BioPandas is build on pandas :P

[a-r-j]

That's a good point! I was mostly concerned about the potential for build problems (mostly as cpdb is my first time working with Cython). I'll make a PR tonight and push a dev release so we can collect some feedback.

[Ruibin-Liu]

One difference in the comparison is that your Cython implementation only reads ATOM, HETATM, and ENDMDL lines while biopandas reads all. Would be interesting to compare the performance if all lines are read (no need to parse like biopandas?).

[a-r-j]

@Ruibin-Liu Hmm, that's a really great point. I could add a read_header arg to cpdb. In any case, I wouldn't have thought it would make a huge difference to speed; in terms of line count PDB files are most coordinates.

[BartoszJanuszNA]

Is this thread dead? I really would like to see it happen, we are looking for something fast to replace BioPython for PDB parsing. I also love Biopandas for its familiar interface so this looks like hitting two birds with one stone :)

[a-r-j]

@BartoszJanuszNA I don't have time to work on this right now but I think it'd be straightforward if you want to pick it up.