search

r3fang / SnapATAC

Analysis Pipeline for Single Cell ATAC-seq
GNU General Public License v3.0
300 stars 125 forks source link

UMAP and t-SNE #51

Open sean00002 opened 5 years ago

sean00002 commented 5 years ago

Hello Rongxin,

I just made a t-SNE plot(Figure1) and want to see what does UMAP look like. But in this case, the UMAP plot(Figure2) looks very weird. Are there any other things I need to make change other than use "umap" instead of "Rtsne" and "tsne"? Also, is there a way I can pull out the barcode of a cluster other than just the cluster index? Thank you for your time!

Best, Sean

x.sp = runViz( obj=x.sp, tmp.folder=tempdir(), dims=2, pca.dims=2:30, weight.by.sd=TRUE, method="Rtsne", fast_tsne_path=NULL, Y.init=NULL, seed.use=10, num.cores=5 )

plotViz( obj=x.sp, method="tsne", point.size=0.5, point.shape=19, point.alpha=0.8, point.color="cluster", text.add=FALSE, text.size=1.5, text.color="black", text.halo.add=TRUE, text.halo.color="white", text.halo.width=0.2, down.sample=10000, pdf.file.name=NULL, pdf.width=7, pdf.height=7, legend.add=TRUE )

Figure1 tsne_snapATAC

Figure2

000026

r3fang commented 5 years ago

hi sorry for the late reply. you can try to use "cosine" distance matrix as input for umap to avoid outliers

TheSallyGardens commented 4 years ago

hi sorry for the late reply. you can try to use "cosine" distance matrix as input for umap to avoid outliers Hi! I have the same question.How to achieve it ？