search

hemberg-lab / SC3

A tool for the unsupervised clustering of cells from single cell RNA-Seq experiments
http://bioconductor.org/packages/SC3
GNU General Public License v3.0
119 stars 55 forks source link

Add and explain non-interactive analysis #21

Closed wikiselev closed 7 years ago

wikiselev commented 8 years ago
  1. Integrate with SCESet class of Bioconductor
  2. Update vignettes and explain how to perform the analysis without interactive session
  3. Add more names and description to the object returned by sc3(interactivity = FALSE)
yanwu2014 commented 8 years ago

I was about to ask if you guys were planning on adding more non-interactive options. It'd be really convenient for me to be able to run the entire process in a script since I'd love to use SC3 as part of a larger pipeline!

yanwu2014 commented 8 years ago

Quick question: is there any way to generate the consensus distance matrix without going through the interactive browser?

Thanks! Yan

wikiselev commented 8 years ago

Hi Yan,

Ok I will start it here, but will try to add a more detailed description to the package vignette very soon.

Hope this helps:

library(SC3)
# run SC3 with interactivity = FALSE
sc3(treutlein, 3:7, interactivity = FALSE)
# all results used in the interactive session are stored in sc3.interactive.arg$cons.table
d <- sc3.interactive.arg$cons.table
# d is a matrix of SC3 results 
# each row corresponds to a combination 
# of checkboxes and the number of clusters k
# in the interactive tool

# now, if you want results from
# "euclidean + pearson + spearman" with "PCA + Spectral" for 5 clusters:
res <- d[d[,1] == "euclidean pearson spearman" & d[,2] == "PCA Spectral" & d[,3] == "5"]

# res is a list in which res[[4]] contains all clustering results for
# the combination of parameters defined in the previous step
clust.res <- res[[4]]

# clust.res is a list again
# clust.res[[1]] - is an unclustered consensus matrix
# clust.res[[2]] - are the cell index labels
# clust.res[[3]] - is an hclust object, corresponding to heirarchical clustering
# of the consensus matrix
# clust.res[[4]] - silhouette index results
consens <- clust.res[[1]]
hc <- clust.res[[3]]

# this command will plot the same plot that you see in the interactive session
library(pheatmap)
pheatmap(
    consens, 
    cluster_rows = hc, 
    cluster_cols = hc, 
    cutree_rows = 5, 
    cutree_cols = 5
)

# get the ordering of the cells in the consensus matrix
ordering <- order.dendrogram(as.dendrogram(hc))

# order cells in the consensus matrix
consens.ordered <- consens[ordering, ordering]

# now you do not need to cluster rows and cols, because consens is already ordered
pheatmap(
    consens.ordered,
    cluster_rows = FALSE, 
    cluster_cols = FALSE 
)

# to get cluster labels
clusts <- cutree(hc, 5)
yanwu2014 commented 8 years ago

Thanks so much! This is perfect for what I'm working on.

Best, Yan

wikiselev commented 7 years ago

The newest version 1.1.7 has a full non-interactive support via scater.

Please follow the manual here: https://cdn.rawgit.com/hemberg-lab/SC3/master/vignettes/my-vignette.html

To be able to use it at the moment you need to reinstall both scater and SC3 from the GitHub: install.packages("devtools") devtools::install_github("davismcc/scater", build_vignettes = TRUE) devtools::install_github("hemberg-lab/SC3")

These features will become available on Bioconductor after the next release of Bioconductor.