/clonealign

Bayesian inference of clone-specific gene expression estimates by integrating single-cell RNA-seq and single-cell DNA-seq data

Primary LanguageRApache License 2.0Apache-2.0

clonealign

Build Status DOI

clonealign assigns single-cell RNA-seq expression to cancer clones by probabilistically mapping RNA-seq to clone-specific copy number profiles using reparametrization gradient variational inference. This is particularly useful when clones have been inferred using ultra-shallow single-cell DNA-seq meaning SNV analysis is not possible.

Version 2.0

Clonealign version 2.0 comes with several updated modelling features. In particular:

  • A multinomial likelihood that vastly increases runtime and removes the need for custom size factors
  • Multiple restarts through the run_clonealign function, where the final fit is chosen as that which maximizes the ELBO

For more info see the NEWS.md file.

Getting started

Vignettes

  1. Introduction to clonealign Overview of clonealign including data preparation, model fitting, plotting results, and advanced inference control
  2. Preparing copy number data for input to clonealign Instructions for taking region/range specific copy number profiles and converting them to gene and clone specific copy numbers for input to clonealign

Installation

clonealign is built using Google's Tensorflow so requires installation of the R package tensorflow. The versioning of Tensorflow and Tensorflow probability currently breaks the standard installation, so the following steps must be taken:

install.packages("tensorflow")
tensorflow::install_tensorflow(extra_packages ="tensorflow-probability", version="2.1.0")
install.packages("devtools") # If not already installed
install_github("kieranrcampbell/clonealign")

Usage

clonealign accepts either a cell-by-gene matrix of raw counts or a SingleCellExperiment with a counts assay as gene expression input. It also requires a gene-by-clone matrix or data.frame corresponding to the copy number of each gene in each clone. The cells are then assigned to their clones by calling

cal <- clonealign(gene_expression_data, # matrix or SingleCellExperiment
                  copy_number_data)     # matrix or data.frame
print(cal)
A clonealign_fit for 200 cells, 100 genes, and 3 clones
To access clone assignments, call x$clone
To access ML parameter estimates, call x$ml_params

print(head(cal$clone))
[1] "B" "C" "C" "B" "C" "B"

Paper

clonealign: statistical integration of independent single-cell RNA and DNA sequencing data from human cancers, Genome Biology 2019

Authors

Kieran R Campbell, University of British Columbia