Fix max_seqlets_subsample bug

[lhqing]

Hi, thank for the great tool! I found a potential bug here.

The merging_max_seqlets_subsample=300 in main function doesn't pass into the SimilarPatternsCollapser, which cause the pattern merge step still run under max_seqlets_subsample=1000

[caenrigen]

@jmschrei @bytewife I am wondering if you spot this PR? and if there is any relevant problems that this bug might create

I suggest to additionally change the default argument from merging_max_seqlets_subsample=300 to merging_max_seqlets_subsample=1000 such that the bug is fixed and the behaviour of modiscolite remains the same as it used to be (when using default values ofc). Note that there is also a notebook that maybe should be fixed too, it uses SimilarPatternsCollapser with max_seqlets_subsample = 300

[EDIT: ] Suggested changes over here caedc9a

[jmschrei]

Thanks for pointing out this issue. I don't think we have time to consider changes like this right now. Potentially after the next set of ENCODE papers comes out there will be more time. Sorry about that.

[caenrigen]

I understand, thank you for the reply 🙂

[austintwang]

Hi @caenrigen, tfmodisco-lite has been merged into the official TF-MoDISco repository. We encourage you to re-open your PR there.