If I calculate cell cycle scores twice, I do not get the same results. Minimal example

Resolved by <a aria-label="Pull request #22" class="issue-link js-issue-link" data-err

Reproducibility issue about pypairs HOT 4 CLOSED

rfechtner commented on July 19, 2024

Reproducibility issue

from pypairs.

Comments (4)

rfechtner commented on July 19, 2024

@fabianrost84 This is expected as the score is calculated by shuffling gene pairs and comparing the actual number matched marker pairs against the shuffled distribution. You can reduce the variance by increasing the iterations parameter when using cyclone(). This makes the distribution more robust but takes longer time to compute.

from pypairs.

fbnrst commented on July 19, 2024

Thanks for the explanation. I can see the shuffling is done here:

pypairs/pypairs/tools/cyclone.py

Line 227 in 8bc94e3

np.random.shuffle(sample_used)

However, I think, it would be great if pypairs would set a random seed such that the result is reproducible. Would you agree? Then, I might open a PR.

from pypairs.

rfechtner commented on July 19, 2024

Seems like a good idea! Makes it easier to share results.

from pypairs.

fbnrst commented on July 19, 2024

Resolved by #22. Thanks @rfechtner!

from pypairs.

Recommend Projects