Tutorial

How to generate synthetic CRISPR data using DKOsimR?

Abbreviations

KO, knockout; SKO, single knockout; DKO, double knockout; %, percentage; GI, genetic interaction; std. dev., standard deviation.

Introduction

DKOsimR is an R package designed for generating synthetic CRISPR double-knockout screening data. It allows researchers to simulate cell growth dynamics and genetic interactions between gene pairs under controlled library setup and experimental conditions.

This tutorial demonstrates:

Installation guide
Graphical overview of simulation framework
List of tunable parameters
Quickstart on default workflow to run simulation
Guide to customize simulation setup and approximate lab data
Summary on why use DKOsimR

Installation

To start running simulation, simply download and install R/RStudio as the first step. You may then install DKOsimR with following commands:

if(!requireNamespace("devtools", quietly = TRUE))
    install.packages("devtools")

devtools::install_github("yuegu-phd/DKOsimR", quiet = TRUE)
devtools::install(dependencies = TRUE)

Make sure all required dependencies are installed using devtools::install(dependencies = TRUE).

Then you may simply load the package:

library(DKOsimR)

Graphical Overview of Study Design

Graphical overview of DKOsim study design

List of Tunable Parameters

Initialized Library Parameters

sample_name: name of the simulation run
coverage: cell representation per guide
n: number of unique single gene
n_guide_g: number of guide per gene
moi: multiplicity of infection - % of cells that are transfected by any virus
sd_freq0: dispersion of initial counts distribution

GI Parameters

p_gi : proportion of interacting gene pairs
sd_gi : std. dev. of re-sampled phenotype with GI presence

Gene Class Parameters

% of theoretical phenotype to each gene class

pt_neg: % negative

pt_pos: % positive

pt_wt: % wild-type

pt_ctrl: % non-targeting control

Mean and std. dev. of theoretical phenotype

mu_neg: mean of negative genes

sd_neg: std. dev. of negative genes

mu_pos: mean of positive genes

sd_pos: std. dev. of positive genes

sd_wt: std. dev. of wild-type genes

Guide Parameters

High-efficacy guides proportion and CRISPR mode

p_high : proportion of high-efficacy guides

mode: CRISPR mode:

use CRISPRn-100%Eff if need 100% effcient guides without randomization

use CRISPRn if need high-efficient guides drawn from distribution

Mean and std. dev. of guide-efficacy

mu_high: mean of high-efficacy guides

sd_high: std. dev of high-efficacy guides

mu_low: mean of low-efficacy guides

sd_low: std. dev of low-efficacy guides

Cell Doublings Parameters

size.bottleneck: bottleneck size - threshold indicating the ceiling of cell growth

n.bottlenecks: number of bottleneck encounters - how many times do we encountering bottlenecks?

n.iterations: number of maximum doubling cycles, by default, we assume a maximum of 30 doublings if we didn’t encounter bottleneck

Randomization Parameter

rseed: values used for random number generator - use same number to control same sets of genes having GI

Miscellaneous

path: path to directory to save outputs of data and logs from simulation

cores_free: number of cores that are left to be free in parallel computing

Quick Start

After loading DKOsimR, to run a simulation with default parameters, you may simply use

dkosim(sample_name = "test", n = 40)

Adjust sample_name and n to name run and initialize number of perturbed genes. Output data will be generated in current working directory.

Alternatively, you may run a simulation in lab approximating mode, by default

dkosim_lab(sample_name = "test_lab", n = 20)

This function applies parameter settings that approximate realistic laboratory data distributions.

Customized Simulation

All tunable parameters may be adjusted by desires in both mode. For example,

dkosim(sample_name="test",
       coverage=10,
       n=60,
       n_guide_g=2,
       sd_freq0 = 1/3.29,
       moi = 0.3,
       p_gi=0.03,
       sd_gi=1.5,
       p_high=1,
       mode="CRISPRn-100%Eff",
       pt_neg=0.15,
       pt_pos=0.05,
       pt_wt=0.75,
       pt_ctrl=0.05,
       mu_neg=-0.75,
       sd_neg=0.1,
       mu_pos=0.75,
       sd_pos=0.1,
       sd_wt=0.25,
       size.bottleneck = 2,
       n.bottlenecks= 1,
       n.iterations = 30,
       rseed = 111,
       path = ".")

Output data will be generated in current working directory.

Simulation Approximating Laboratory Data

DKOsimR also provides a wrapper function for lab approximating mode to simulate data that resembles real laboratory CRISPR screening datasets:

dkosim_lab(sample_name = "test_lab", n = 20)

This function applies parameter settings that approximate realistic laboratory data distributions.

All parameters can be further customized by users to fit specific experimental setup as desired in both mode, for example:

dkosim(sample_name="test",
       coverage=10,
       n=60,
       n_guide_g=2,
       sd_freq0 = 1/3.29,
       moi = 0.3,
       p_gi=0.03,
       sd_gi=1.5,
       p_high=1,
       mode="CRISPRn-100%Eff",
       pt_neg=0.15,
       pt_pos=0.05,
       pt_wt=0.75,
       pt_ctrl=0.05,
       mu_neg=-0.75,
       sd_neg=0.1,
       mu_pos=0.75,
       sd_pos=0.1,
       sd_wt=0.25,
       size.bottleneck = 2,
       n.bottlenecks= 1,
       n.iterations = 30,
       rseed = 111,
       path = ".")

Summary

DKOsimR enables researchers to:

generate reproducible synthetic CRISPR screening datasets
benchmark genetic interaction detection methods
evaluate and optimize experimental design parameters

For further information, please refer to the API documentation and the vignettes file (PDF) of DKOsimR R package.