ScreenPro2 0.3.1

Flexible analysis of high-content CRISPR screening

ScreenPro2

TL;DR

ReadTheDocs | PyPI

ScreenPro2 enables perform flexible analysis on high-content CRISPR screening datasets. It has functionalities to process data from diverse CRISPR screen platforms and is designed to be modular to enable easy extension to custom CRISPR screen platforms or other commonly used platforms in addition to the ones currently implemented.

---

Introduction

Functional genomics field is evolving rapidly and many more CRISPR screen platforms are now developed. Therefore, it's important to have a standardized workflow to analyze the data from these screens. ScreenPro2 is provided to enable researchers to easily process and analyze data from CRISPR screens. Currently, you need to have a basic background in programming (especially Python) to use ScreenPro2.

ScreenPro2 is conceptually similar to the ScreenProcessing pipeline but ScreenPro2 is designed to be more modular, flexible, and extensible. Common CRISPR screen methods that we have implemented here are illustrated in a recent review paper:

From: A new era in functional genomics screens

Fig. 1: Common types of CRISPR screening modalities indicating advances in CRISPR methods.

Installation

ScreenPro2 is available on PyPI and can be installed with pip:

pip install ScreenPro2

---

For the latest version (development version) install from GitHub:

pip install git+https://github.com/ArcInstitute/ScreenPro2.git

Usage

First, import the ScreenPro2 package:

import screenpro as scp

Data analysis for CRISPR screens with NGS readouts can be broken down into three main steps:

• Step 1: FASTQ to counts
• Step 2: Phenotype calculation
• Step 3: Explore results and QC reports

Step 1: FASTQ to counts

ScreenPro2 has a built-in method to process FASTQ files and generate counts. This method is implemented in the ngs module and relvent submodules. A minor novelty here has enabled processing single, dual, or multiple sgRNA CRISPR screens. Also, this approach can retain recombination events which can occur in dual or higher order sgRNA CRISPR screens.

Currently, Counter class from the ngs module can process FASTQ files and generate counts for standard CRISPR screens with single or dual guide design.

Here is a draft code to process FASTQ files and generate counts for an experiment with CRISPRi-dual-sgRNA-screens:

# Initialize the Counter object
counter = scp.Counter(cas_type = 'cas9', library_type = 'single_guide_design')

# Load the reference library
counter.load_library("<path-to-CRISPR-library-table>", sep = '\t', verbose = True, index_col=None)

# Define the samples
samples = [] 
## `samples` is a list of sample ids in the experiment. 
## Each sample id should match the sample name in the FASTQ files, i.e. <sample_id>.fastq.gz

# Process the FASTQ files and generate counts
counter.get_counts_matrix(
    fastq_dir = '<path-to-fastq-directory>',
    samples = samples,
    verbose = True
)

Here is a draft code to process FASTQ files and generate counts for an experiment with CRISPRi-dual-sgRNA-screens:

# Initialize the Counter object
counter = scp.Counter(cas_type = 'dCas9', library_type = 'dual_guide_design')

# Load the reference library
counter.load_library("<path-to-CRISPR-library-table>", sep = '\t', verbose = True, index_col=None)

# Define the samples
samples = []
## `samples` is a list of sample ids in the experiment.
## Each sample id should match the sample name in the FASTQ files, i.e. <sample_id>_R[1,2].fastq.gz

# Process the FASTQ files and generate counts
counter.get_counts_matrix(
    fastq_dir = '<path-to-fastq-directory>',
    samples = samples,
    verbose = True
)

After this, you have .counts_mat calculated in the Counter object.

---

To proceed, you need to create an AnnData object from the counts matrix and metadata. You can use the following code to create an AnnData object:

adata = counter.build_counts_anndata()

Step 2: Phenotype calculation

Once you have the counts, you can use ScreenPro2 phenoscore and phenostats modules to calculate the phenotype scores and statistics between screen arms.

Load Data

First, load your data into an AnnData object (see anndata for more information).

The AnnData object must have the following contents:

• adata.X – counts matrix (samples x targets) where each value represents the sequencing count from NGS data.
• adata.obs – a pandas dataframe of samples metadata including "condition" and "replicate" columns.
  • "condition": the condition for each sample in the experiment.
  • "replicate": the replicate number for each sample in the experiment.
• adata.var – a pandas dataframe of targets in sgRNA library including "target" and "targetType" columns.
  • "target": the target for each entry in reference sgRNA library. For single sgRNA libraries, this column can be used to store gene names. For dual or multiple targeting sgRNA libraries, this column can be used to store gene pairs or any other relevant information about the target.
  • "targetType": the type of target for each entry in reference sgRNA library. Note that this column is used to distinguish between different types of sgRNAs in the library and negative control sgRNAs can be defined as "targetType" == "negCtrl". This is important for the phenotype calculation step.

ScreenPro2 has a built-in class for different types of CRISPR screen assays. Currently, there is a class called PooledScreens that can be used to process data from pooled CRISPR screens. To create a PooledScreens object from an AnnData object, you can use the following example code:

import pandas as pd
import anndata as ad
from screenpro.assays import PooledScreens

adata = ad.AnnData(
    X   = counts_df, # pandas dataframe of counts (samples x targets)
    obs = meta_df,   # pandas dataframe of samples metadata including "condition" and "replicate" columns
    var = target_df  # pandas dataframe of targets metadata including "target" and "targetType" columns
)

screen = PooledScreens(adata)

Perform Screen Processing Analysis

Once the screen object is created, you can use several available workflows to calculate the phenotype scores and statisitics by comparing each entry in reference sgRNA library between screen arms. Then, these scores and statistics are used to nominate hits.

Drug Screen Workflow: calculate gamma, rho, and tau scores

.calculateDrugScreen method can be used to calculate the enrichment of each gene between screen arms for a drug screen experiment. This method calculates gamma, rho, and tau scores for each gene and adds them to the .phenotypes attribute of the PooledScreens object.

Here is an example for running the workflow on a CRISPRi-dual-sgRNA-screens dataset:

# Run the ScreenPro2 workflow for CRISPRi-dual-sgRNA-screens
screen.calculateDrugScreen(
  t0='T0',
  untreated='DMSO',  # replace with the label for untreated condition
  treated='Drug',    # replace with the label for treated condition
  db_untreated=1,    # replace with doubling rate of untreated condition
  db_treated=1,      # replace with doubling rate of treated condition
  score_level='compare_reps'
)

---

For example, in a Decitabine CRISPRi drug screen (see Figure 1B-C in this bioRxiv paper), each phenotype score represents a comparison between different arms of the screen and rho scores shows the main drug phenotype as illustrated here:

Flow cytometry based screen workflow: calculate phenotype score to compare high and low bins

.calculateFlowBasedScreen method can be used to calculate the enrichment of each target between high bin vs. low bin of a flow cytometry-based screen experiment. This method calculates PhenoScore for each target and adds them to the .phenotypes attribute of the ScreenPro object.

# Run the ScreenPro2 workflow for CRISPRi-dual-sgRNA-screens
screen.calculateFlowBasedScreen(
  low_bin='low_bin', high_bin='high_bin',
  score_level='compare_reps'
)

Step 3: Explore results and QC reports

Once the phenotypes are calculated, you can extract and explore the results using the .phenotypes attribute of the ScreenPro object. Currently, there are very limited functionalities built-in to visualize the results, but we are working on adding more features to make it easier for users. However, you can easily extract the results and use other libraries like seaborn and matplotlib in Python or ggplot2 in R to visualize the results.

---

Supported CRISPR Screen Platforms

One of the main goals of ScreenPro2 is to make it easy to process data from commonly used CRISPR screen platforms. Also, it is designed to be modular to enable easy extension to custom CRISPR screen platforms or other commonly used platforms in addition to the ones currently implemented.

---

Currently, ScreenPro2 has easy-to-use workflows for the following CRISPR screen platforms:

dCas9 CRISPRa/i single-sgRNA screens

Horlbeck et al., eLife (2016)

Horlbeck et al. developed a CRISPR interference (CRISPRi) and CRISPR activation (CRISPRa) screening platform that uses a single sgRNA within a single plasmid and then there are up to 10 sgRNAs per gene. The multiple sgRNAs per gene can be used to perfrom statistical comparisons in guide-level or gene-level between screen arms. ScreenProcessing has been developed to process data from this type of screen. We reimplemented the same workflow in ScreenPro2 and it has all the necessary tools to process data from this type of screen.

dCas9 CRISPRa/i dual-sgRNA screens

Replogle et al., eLife (2022)

Replogle et al. developed a CRISPR interference (CRISPRi) and CRISPR activation (CRISPRa) screening platform that uses two sgRNAs per gene within a single plasmid, and it has been used to perform genome-scale CRISPRi screens. ScreenPro2 has all the necessary tools to process data from this type of screen.

License

ScreenPro2 is licensed under the terms of the MIT license (see LICENSE for more information) and developed by Abolfazl (Abe) Arab (@abearab) as a Research Associate in the Gilbert lab at UCSF and Arc Institute.

Citation

If you use ScreenPro2 in your research, please cite the following paper.

Coming soon...