CRISPR KO Analysis based on Genomic Editing data —— Toward personalized sgRNA design in heterogeneous experimental conditions

A CRISPR-cas9 based genome-editing data analysis resource and platform, for the analysis of indels and microhomology patterns, the identification of personalized features correlated to sgRNA KO efficiency on heterogeneous experimental conditions, and the evaluation of the sgRNA KO efficiency based on the CRISPR-Cas9 KO NGS data or the sgRNA KO assay data.

The ultimate goals of CAGE:

1. CAGE provides a standard CROWDSOURCING platform for users to share the CRISPR-Cas9 based gene KO data.

2. CAGE provides an efficient interface to analyze and visualize the CRISPR-based KO NGS data.

3. CAGE provides a robust learning pipeline to derive the sequence determinants from heterogeneous genome editing data.

4. CAGE provides an personalized evaluation framework for on-target sgRNA design based on the derived sequence determinants torward specific cell types or organisms.

Currently CAGE records the optimal sgRNA KO efficiency prediction models and the personalized evaluation models in sgRNA design for the following different cell types. The optimal results for new cell types as well as the the current ones will be updated regularly. Users can select the existing evaluation model for a specific cell type for sgRNA design, or they can use their own sgRNA KO data to generate a new personalized evaluation model for a new cell type for furthur sgRNA KO efficiency evaluation.

• Python 2.7
• Numpy >= 1.9.2
• Scipy >= 0.15.1
• Pandas >= 0.16.0
• scikit-learn >= 0.16.1
• lxml >= 3.4.4
• pyfasta >= 0.5.2
• bwa >= 0.7.12
• samtools >= 0.1.19
• bedtools >= 2.23.0
• pyslep (for multi-task group lasso)
• LaTeX (for visualization)

Make sure to perform this presetting carefully. Because reference setting is very important.

For the sake of simplicity, we use hg19 as the example.

1. Download the hg19 genome(fasta file) from UCSC, put it in certain directory, name it hg19.fa and set the directory path as $FASTADB.

2. Generate bwa index files from hg19.fa, put them in certain directory and set the directory path as $BWADB.

3. (Optional) Download the hg19 gene annotation files from UCSC, convert it to bed-6 format with the 4th column being the gene name, put them in certain directory and set the directory path as $BEDDB. Here are the renamed file:

git clone https://github.com/bm2-lab/cage-dev.git

For performing multi-task group lasso, pyslep is necessary.

1. cd pyslep
2. sh setup_pyslep.sh

python cage.py <command> [option] ...

1. sg Process sgRNA sequences into sgRNA information table
2. prep Process NGS data
3. mh Microhomology Detection
4. indel Indel frameshifting paradigm analysis
5. fs Feature selection and model prediction on clearly defined sgRNA KO efficiency
6. mt Feature selection with multi-task group LASSO on clearly defined sgRNA KO efficiency for cross-platform data
7. eval sgRNA KO efficiency evaluation and the scanning of a given genome region for sgRNA design
8. vis Visualization of feature selection result

#####Note For label file, the first and the last column will be regarded as sgID and score respectively. File header should exist and the header of the first column must be sgID. See the following example.

Generate sgRNA Information Table (sg file)

python cage.py sg -s <sgRNA.fq>
	              -o <output directory>
                  -g <reference genome> (e.g. hg19)
				  -t <bwa threads> (default 1)
				  -a (annotate, optional)

For more detail on the options, see python cage.py sg -h.

• Single-end

python cage.py prep -s <sg file>
	                -f <reads.fq>
	                -o <output directory>
                    -g <reference genome>
					-t <bwa threads>

• Paired-end

python cage.py prep -s <sg file>
                    -f <reads_1.fq>
					-r <reads_2.fq>
					-o <output directory>
					-g <reference genome>
					-t <bwa threads>

For more detail on the options, see python cage.py prep -h.

python cage.py mh -i <samind file>
                  -o <output directory>
	              -g <reference genome>

For more detail on the options, see python cage.py mh -h.

Generate sgRNA-indel table (iost file)

python cage.py indel -i <samind file>
                     -s <sg file>
                     -o <output directory>
	                 -g <reference genome>
					 -t <read-count cutoff> (default 0)

For more detail on the options, see python cage.py indel -h.

• Manual

python cage.py fs -i <label file>
                  -s <sg file>
                  -o <output directory>
	              -g <reference genome>
				  -t <reads cutoff> (default 0)
				  -u <upstream region length> (default 30)
				  -w <downstream region length> (without PAM, default 27)
				  -c <cross-validation folds> (default 5)
				  -j <number of CPU cores used> (default 1)
				  -m <lasso|logit>

• Auto

python cage.py fs -i <label file>
                  -s <sg file>
                  -o <output directory>
	              -g <reference genome>
				  -t <reads cutoff> (default 0)
				  -a (auto detection for sequence region)
				  --init-radius <init radius> (default 0)
				  -r <radius> (default 200)
				  --step <detection step> (default 5)
				  -c <cross-validation folds> (default 5)
				  -j <number of CPU cores used> (default 1)
				  -m <lasso|logit> (method)

For more detail on the options, see python cage.py fs -h.

python cage.py mt -i [<label file> [<label file> ...]]
                  -s [<sg file> [<sg file> ...]]
                  -o <output directory>
	              -g [<ref genome> [<ref genome> ...]]
				  -u <upstream region length> (default 30)
				  -w <downstream region length> (without PAM, default 27)
				  -d <selection strictness>

• Evaluation with Genome Scanner

python cage.py eval -c <target chromosome>
                    -b <start coordinate>
                    -e <end coordinate>
                    -m <pkl file>
                    -o <output directory>
					-g <reference genome>
					-d <two-sided|pos|neg> (scan direction)
					-t <bwa threads>

• Evaluation with sgRNA Information Table

python cage.py eval -s <sg file>
                    -m <pkl file>
                    -o <output directory>
					-g <reference genome>

For more detail on the options, see python cage.py eval -h.

python cage.py vis -f <feature report file>
                   -o <output directory>

For more detail on the options, see python cage.py vis -h.

To run examples, cd example first, then execute the following commands.

• sg: sh exam.sh sg
• prep for single-end: sh exam.sh prep_se
• prep for pair-end: sh exam.sh prep_pe
• mh: sh exam.sh mh
• indel: sh exam.sh indel
• fs using LASSO without auto detection: sh exam.sh fs_las
• fs using LASSO with auto detection: sh exam.sh fs_las_a
• fs using Logistic Regression without auto detection: sh exam.sh fs_log
• fs using Logistic Regression with auto detection: sh exam.sh fs_log_a
• mt: sh exam.sh mt
• eval using genome scanner: sh exam.sh eval
• eval using sgRNA information table: sh exam.sh eval_sg
• vis: sh exam.sh vis