About

Single-cell copy number calling and event history reconstruction.

A statistical model and MCMC algorithm tailored to single-cell copy number profiling from shallow whole-genome DNA sequencing data. SCICoNE reconstructs the history of copy number events in the tumour and uses these evolutionary relationships to identify the copy number profiles of the individual cells.

Requirements

• C++ compiler that supports C++14 standards (e.g. gcc>=5.2.0, clang>=5.0.0))
• CMake >= 3.9
• Boost >= 1.6.x
• OpenMP >= 4.5
• NLopt >= 2.6.2

Installation

Once the requirements are in place, downloading and installing SCICoNE takes about 5 minutes.

git clone git@github.com:cbg-ethz/SCICoNE.git # Clone the repository
cd SCICoNE
mkdir build && cd build                       # Create and enter the build directory
cmake ..                                      # Compile the program with cmake
make                                          # Build the executables

Python wrapper

We also provide a Python 3 wrapper to facilitate plotting and easily integrate SCICoNE with other data analysis tools. This interface runs the C++ binaries and reads the outputs into numpy arrays.

cd pyscicone
pip install .

Quick start

SCICoNE takes a read counts matrix of cells by genomic bins and outputs the copy number profile of each cell and the underlying event history. Two introductory notebooks showcasing the SCICoNE workflow and API are available:

• Tutorial
• Example run on 10X Genomics data

C++ reference

Breakpoint detection

Find the breakpoints defining copy number regions from a cells by bins matrix.

Parameter name	Description	Default value
d_matrix_file	Path to the counts matrix file, delimiter: ' ', line separator: '\n'	""
n_cells	Number of cells in the input matrix	-
n_bins	Number of bins in the input matrix	-
min_cells	Minimum number of cells to consider for a bin being a breakpoint	4
window_size	Size of the window used in breakpoint detection	10
threshold	Breakpoint calling sensitivity. If -1, stop after computing the LR	3
bp_min	Minimum number of breakpoints	0
bp_limit	Maximum number of breakpoints	300
compute_lr	Boolean indicator of wether the per bin cell-wise breakpoint evidence should be computed	True
lr_file	Path to a matrix containing the evidence for breakpoint at each bin in each cell	""
sp_file	Path to a vector containing the combined evidence for breakpoint at each bin across all cells	""
compute_sp	Boolean indicator of wether the per bin breakpoint evidence should be computed	True
evaluate_peaks	Boolean indicator of wether to evaluate peaks and call breakpoints	True
input_breakpoints_file	Path to file indicating bins which correspond to known breakpoints that must be included. These may be, for example, the chromosome bin coordinates.	""
verbosity	Verbosity of the programme, 0 is non-verbose setting, 1 creates the debug files, 2 writes the logs as well	0
postfix	Postfix to be added to the output files, this is useful when you are running multiple simulations through a workflow management system	""
print_precision	The precision points of the score values to be printed	16

Sample run :

$ ./breakpoint_detection --d_matrix_file ./d_mat.txt --n_bins 1000 --n_cells 400 --window_size 10 --threshold 3.0 --bp_limit 300 --verbosity=1 --evaluate_peaks=True

Inference

Finds the maximum a posteriori tree given a cells by regions matrix.

Parameter name	Description	Default value
region_sizes_file	Path to the file containing the region sizes, each line contains one region size. Segmentation is performed if the region sizes file is not specified	""
d_matrix_file	Path to the cells by regions counts matrix file, delimiter: ' ', line separator: '\n'	""
n_iters	Number of iterations	10000
n_cells	Number of cells in the input matrix	-
n_regions	Number of regions in the input matrix	-
ploidy	The ploidy information	2 (diploid, human)
verbosity	Verbosity of the programme, 0 is non-verbose setting, 1 creates the debug files, 2 writes the inference logs as well, 3 writes the tree logs on top	0
seed	Seed	-
postfix	Postfix to be added to the output files, this is useful when you are running multiple simulations through a workflow management system	""
print_precision	The precision points of the score values to be printed	16
----	parameters for the random initialised tree	---
n_nodes	the number of nodes in the random initialised tree	50
lambda_r	lambda param for the poisson that generates the number of regions	0.1
lambda_c	lambda param for the poisson that generates the copy number state of a region	0.2

Sample run :

$ ./inference --n_cells 400 --n_regions 10 --n_iters 100 --n_nodes 10 --ploidy 2 --verbosity 2 --seed 42 --d_matrix_file ./30_d_mat.txt --region_sizes_file ./30_region_sizes.txt

Simulation

Simulates the count matrix. Outputs the count matrix, region sizes, ground truth CNVs and the tree that generated the data.

Parameter name	Description	Default value
n_bins	Number of bins of the input matrix	1000
n_cells	Number of cells	400
n_nodes	Number of nodes of the tree	50
n_regions	Number of regions	50
n_iters	Number of iterations	10000
n_reads	Number of reads per cell	10000
ploidy	The ploidy information	2 (diploid, human)
verbosity	Verbosity of the programme, 0 is non-verbose setting, 1 creates the debug files, 2 writes the inference logs as well, 3 writes the tree logs on top	0
seed	Seed	-
postfix	Postfix to be added to the output files, this is useful when you are running multiple simulations through a work flow management system	""
print_precision	The precision points of the score values to be printed	16

Sample run :

$ ./simulation --n_cells 400 --n_bins 1000 --n_regions 10 --n_nodes 10 --n_reads 100000 --ploidy 2 --verbosity 2  --seed 42 --print_precision 32

Test

Runs the validation tests and writes the results to the standard output and error streams.

Sample run :

$ ./tests