search

KrishnaswamyLab / SAUCIE

MIT License
98 stars 29 forks source link

Uninteresting results - additional parameters? #7

Open lynnyi opened 5 years ago

lynnyi commented 5 years ago

I ran the default python SAUCIE.py (default settings) on two datasets (with clustering and batch correction), which gave only "0.0" cluster annotation for all cells. The SAUCIE embedding was just a linear embedding of all the cells (pretty much on the y=x line). Are there additional parameters that must be tuned, i.e. the lambdas? What are they?

mattamodio commented 5 years ago

Hmm that could have a couple of different causes, but I think the range of the data is still the most likely culprit. What are the smallest and largest values in your cells(rows) x markers(columns) matrix? After log scaling, hopefully there are no numbers larger than 10-20?

The lambdas will affect the number of clusters identified, but if the embedding is close to the y=x line, that is more likely to come from data being outside of the [-10,10] range.