Closed matsen closed 5 years ago
Just a quick note: I do have moderate experience with pandas (like, 3 or 4 months' worth). But I absolutely agree that the nature of the data should dictate the language.
As of this AM I'm definitely leaning towards Python. The main reason is that all of the other code we have is in Python, and if we want to preserve the vision of having worthwhile interaction between partis, cft, Jean and David's motif code, and this, then we're going to want to use the same language.
I apologize in advance for matplotlib. We can totally use this http://rpy2.readthedocs.io/en/version_2.7.x/graphics.html if you like for graphics that aren't part of some code for export, but rather just for the paper.
I have a working function that uses sumrep to get annotated sequences (using partis), and then constructs a list of mutation indices for each mature/naive pair. The latter part shouldn't be too difficult to redo in python, but do you know if there is a good way to call partis from python? It might not be hard to do using the run_partis.sh script I wrote, but I don't want to reinvent the wheel. Or maybe we can somehow import sumrep into a python module?
Ha! We were thinking in parallel (see #4 ).
I don't know to what extent partis can be used as a library. If Duncan's not around you could check in with David. In fact, they've thought a lot about mutations so chatting with him would be good in general.
I love Python, and find R to be an amazing statistical environment but a strange language for complex programming. However, this is balanced by you all knowing R better than Python.
As we are deciding, I think that the key question is what the data structures we want look like. If they are tabular, then R will be lovely, and you will end up having growing pains with Pandas. If they are not, then Python's the choice for sure.
Assuming that we want to use bowtie2, one question is SAM/BAM support.