This repository (CIN_PDO) contains the simulation program and scripts used for modelling chromosomal instability (CIN) observed in the patient-derived tumor organoid (PDO) data in paper:
Reconstructing single-cell karyotype alterations in colorectal cancer reveals punctuated and gradual diversification patterns. Yannik Bollen, Ellen Stelloo, Petra van Leenen, Myrna van den Bos, Bas Ponsioen, Bingxin Lu, et al. (2021), Nature Genetics, accepted.
The main program (in directory code) simulates the growth of a patient-derived tumor organoid with a stochastic birth-death branching process. It generates cell lineage tree and single-cell copy number profiles (CNPs). It can also simulate bulk CNPs (average CNPs of all cells in an organoid) of multiple organoids. Only relative arm-/chr-level CNAs are simulated. All CNAs are assumed to be reciprocal. A set of summary statistics is computed from simulated data, which can be used by ABC (approximate Bayesian computation) approach to infer mutation rates and selection strengths.
The directory script contains the main scripts used for the modelling (parameter estimation, model selection, power analysis, and likelihood-ratio test) in the paper.
The directory data contains the main data and modelling results in the paper.
Download the source code and then run make in folder "code" to generate the simulation program 'simpdo'.
Please run ./simpdo -h for all options.
Option "verbose" controls how much information to output.
-
When
verbose=0, a vector of summary statistics is written to the standard output. The summary statistics represent:- average absolute unique chr-level copy number changes,
- average absolute unique arm-level copy number changes,
- half lineage tree length,
- average ratios of branch length starting from the parent of a node to that starting from the daughter of the node,
- number of altered arms across all cells (count 1 when one arm is altered in any cell),
- sum of average CN of all cells across all arms,
- number of de novo CNAs,
- number of arms altered in more than one cell.
The first four values were used for ABC inference of mutation rates, birth rates, and selection coefficients.
-
When
verbose=1, two files will be output, including the copy numbers for each cell in the final population and summary information of the simulation. -
When
verbose=2, five additional files will be output, including the copy numbers, mutations and lineage tree for all cells generated in the branching process.
- Assume neutral evolution:
./simpdo -o ./ -e 25 --chr_prob 0.1 --arm_prob 0.1 --birth_rate 0.5 --skip 3 --seed $RANDOM --verbose 0
- Assume selection:
./simpdo -o ./ -e 25 --model 1 --fitness "1 -0.1" --chr_prob 0.1 --arm_prob 0.1 --birth_rate 0.5 --skip 3 --seed $RANDOM --verbose 0
Here, "--model 1" indicates selection is imposed. Option --fitness "1 -0.1" specifies there is one selection coefficient being -0.1, which indicates negative selection.
By default, gradual selection based on parent cell is assumed. When a daughter cell accumulates new CNAs after division, its birth rate will be changed compared with the parent cell.
The simulation of multiple organoids is based on parameters inferred from real single-cell CNPs of cells from individual organoids. The output of ABC inference on individual organoids is required.