monchin
commented
5 years ago To get totally same result is almost impossible, because there are several random process in SPADE. Not only in downsample, but also in generating layout from a minimum spanning tree. You can cancel downsampling by setting the target number the same as the original total number, but I'm afraid it would be very very very slow because the time complexity of agglomerative cluster is large, so I strongly don't recommend to cancel it. k is how many clusters you'd like to get. One cluster is one node ( or one kind of cells ) in the final plots. The default value is 200, but in my opinion it is too large. You need to modify it to find a relatively best value.
hello， I want to analysis my single call rna-seq data with spade package version 1.16。But I can't get a result with good tree structure 。So I download SupplementaryData4.fcs dataset from paper ‘Visualization and cellular hierarchy inference of single-cell data using SPADE’。I want to get the same result in paper。 Command lines are as follows；
SPADE.driver(‘SupplementaryData4.fcs’, out_dir=output_dir, transforms=flowCore::arcsinhTransform(a=0, b=0.2), k=72, downsampling_target_number = NULL, downsampling_target_pctile=1, downsampling_target_percent=1)I tried to cancel downsampling，but I am not sure it was working。And I really confused with the K value。I really need your help。best wish