search

hemberg-lab / scmap

A tool for unsupervised projection of single cell RNA-seq data
http://bioconductor.org/packages/scmap
GNU General Public License v3.0
90 stars 11 forks source link

Feature selection #3

Closed davetang closed 6 years ago

davetang commented 6 years ago

Hi there! Thank you for the very useful package. I have a question on selectFeatures and whether other methods, such as using highly variable genes, will be implemented? I realise that the M3Drop preprint suggests that it out performs other feature selection methods, at least in full-transcript scRNA-seq protocols (I have UMI data by the way). If I want to implement something myself, is it simply defining a logical vector to scmap_features for the genes I want to use?

My problem is that the M3Drop method doesn't identify representative genes that define one of my clusters and hence even when I project the data onto itself, only half of the cells in that cluster are assigned back to that cluster. I've tried increasing the number of features/genes up to 2,000 but it doesn't improve.

wikiselev commented 6 years ago

Hi, thanks for your question! The short answer is: yes, if you don't like the results of M3Drop you can select features manually by defining the scmap_features vector in the rowData slot:

rowData(YOUR_SCE_OBJECT)$scmap_features <- YOUR_VECTOR

scmap_features values must be either TRUE or FALSE corresponding to selected and non-selected genes accordingly. This should be sufficient for your purposes, but please let me know if there are any errors.

davetang commented 6 years ago

Brilliant, thanks!