theislab / scib

Benchmarking analysis of data integration tools

MIT License

294 stars 63 forks source link

runBBKNN returns None if n_obs in [1e5, 2.5e5] #230

Closed jamesbaye closed 3 years ago

jamesbaye commented 3 years ago

Appears to be a bug.

LuckyMD commented 3 years ago

Thanks for pointing that out! We didn't have a dataset between these sizes and therefore must have missed this. Indeed it should be 1e5 for both boundaries. Could you fix this @danielStrobl? I think 1e5 is written in the manuscript at least, but should double check there as well.

danielStrobl commented 3 years ago

231

LuckyMD commented 3 years ago

Just for a heads up though... if you intend to use bbknn, i would use it from the original package and adapt the parameters to your dataset.

jamesbaye commented 3 years ago

Thanks @LuckyMD, I'm running bbknn alongside several other methods to compare with an in-house method we're developing so default parameters accessible with your current API should be enough for my current analysis. Thanks for the resource!

LuckyMD commented 3 years ago

Glad it's useful! We're happy for any feedback to make the package easier to use in case you run into more issues! We're in the process of setting up 3 separate packages to make this one pip installable ideally