findthreshold()

[[e-mail]]

Dear folks!

I run findthreshold function on a dataset of ~30,000 sequences and it takes me several hours -I have no choice but run it overnight. Do you encounter also long running of this function?

Sivan

[Namita Gupta]

Unfortunately, yes, it is very slow. The slow step is in finding the appropriate bandwidth parameter for smoothing the distance-to-nearest distribution. I did check that the results are very robust by subsampling down to 15,000 sequences, but it may be possible to subsample even further, if you want to explore that.

[Jason Vander Heiden]

We've added a new Guassian Mixture Model method to the findThreshold function (method="gmm") which is significantly faster than the old smoothed density estimate. This is now the default.