xqbai / ccnmf

A Coupled-Clone Nonnegative Matrix Factorization Method for Integrating Tumor Single-Cell RNA and DNA Sequencing Data

CCNMF

Overview

Coupled-Clone Nonnegative Matrix Factorization (CCNMF) is a method for joint inference of clonal structure using paired Single-cell DNA-Seq and RNA-Seq data. The framework is based on optimizing an objective function that maximizes clone structure coherence between single-cell gene expression and copy number profiles, in which the two profiles are coupled by the dosage effect. The coupling dosage effect can be estimated prior either by a linear regression model using publicly available paired RNA and DNA bulk sequencing data (TCGA), or by using an uninformative prior. The workflow in CCNMF is illustrated in the figure below.

Installation

install.packages('devtools')
devtools::install_github("labxscut/CCNMF")

Analyze gastric cancer NCI-N87 cell line

Raw data

• Download NCI-N87 single-cell RNA-seq data from Gene Expression Omnibus (GSE142750).

• Download NCI-N87 single-cell DNA-seq FASTQ files from National Institute of Health’s SRA repository PRJNA498809.

• The raw data can be accessed from NCI_N87 raw data.

Preprocess data

• The preprocessing pipeline is as example/NCI-N87/NCI_N87_preprocess.R.

• The processed scDNA-seq and scRNA-seq matrics are available at CCNMF/data/processed_data/NCI_N87.

Run CCNMF

• The pipeline using CCNMF to analyze NCI-N87 cell line datasets is available at CCNMF/example/NCI_N87/Run_CCNMF.R.

---

Analyze primary gastric cancer P5931

Preprocess data

• The processed scDNA-seq and scRNA-seq matrics are available at CCNMF/data/processed_data/P5931.

Run CCNMF

• The pipeline using CCNMF to analyze P5931 is aviailable at CCNMF/example/P5931/Run_CCNMF.R.

---

Simulation

Generate simulated paired scDNA and scRNA

• https://nbviewer.jupyter.org/github/XQBai/CCNMF/blob/master/notebooks/SimulationProcedures.ipynb

Reference

Xiangqi Bai, Zhana Duren, Lin Wan, and Li C. Xia. Joint Inference of Clonal Structure using Single-cell Genome and Transcriptome Sequencing Data NAR Genomics and Bioinformatics

Contact

[email protected] and [email protected].

ccnmf's Issues

Running time too long

Hi,

I tried to apply it to a dataset that has 93 cells with scRNA-seq and 93 cells with CNV, and around 10,000 genes. I have been running CCNMF on this dataset for two weeks, but it hasn't finished yet. Is there any way to accelerate the running process?

Thanks for your help!