zhaoxiangsimoncai / crispy

Python module to analyse CRISPR-based libraries

Module with utility functions to process CRISPR-based screens and method to correct gene independent copy-number effects.

Description

Crispy uses Sklearn implementation of Gaussian Process Regression, fitting each sample independently.

Install

Install pybedtools and then install Crispy

conda install -c bioconda pybedtools

pip install cy

Examples

Support to library imports:

from crispy.CRISPRData import Library

# Master Library, standardised assembly of KosukeYusa V1.1, Avana, Brunello and TKOv3 
# CRISPR-Cas9 libraries.
master_lib = Library.load_library("MasterLib_v1.csv.gz")


# Genome-wide minimal CRISPR-Cas9 library. 
minimal_lib = Library.load_library("MinLibCas9.csv.gz")

# Some of the most broadly adopted CRISPR-Cas9 libraries:
# 'Avana_v1.csv.gz', 'Brunello_v1.csv.gz', 'GeCKO_v2.csv.gz', 'Manjunath_Wu_v1.csv.gz', 
# 'TKOv3.csv.gz', 'Yusa_v1.1.csv.gz'
brunello_lib = Library.load_library("Brunello_v1.csv.gz")

Select sgRNAs (across multiple CRISPR-Cas9 libraries) for a given gene:

from crispy.GuideSelection import GuideSelection

# sgRNA selection class
gselection = GuideSelection()

# Select 5 optimal sgRNAs for MCL1 across multiple libraries 
gene_guides = gselection.select_sgrnas(
    "MCL1", n_guides=5, offtarget=[1, 0], jacks_thres=1, ruleset2_thres=.4
)

# Perform different rounds of sgRNA selection with increasingly relaxed efficiency thresholds 
gene_guides = gselection.selection_rounds("TRIM49", n_guides=5, do_amber_round=True, do_red_round=True)

Copy-number correction:

import crispy as cy
import matplotlib.pyplot as plt

# Import data
rawcounts, copynumber = cy.Utils.get_example_data()

# Import CRISPR-Cas9 library
lib = cy.Utils.get_crispr_lib()

# Instantiate Crispy
crispy = cy.Crispy(
    raw_counts=rawcounts, copy_number=copynumber, library=lib
)

# Fold-changes and correction integrated funciton.
# Output is a modified/expanded BED formated data-frame with sgRNA and segments information
bed_df = crispy.correct(x_features='ratio', y_feature='fold_change')
print(bed_df.head())

# Gaussian Process Regression is stored
crispy.gpr.plot(x_feature='ratio', y_feature='fold_change')
plt.show()

Credits and License

Developed at the Wellcome Sanger Institue (2017-2020).

For citation please refer to:

Gonçalves E, Behan FM, Louzada S, Arnol D, Stronach EA, Yang F, Yusa K, Stegle O, Iorio F, Garnett MJ (2019) Structural rearrangements generate cell-specific, gene-independent CRISPR-Cas9 loss of fitness effects. Genome Biol 20: 27

Gonçalves E, Thomas M, Behan FM, Picco G, Pacini C, Allen F, Parry-Smith D, Iorio F, Parts L, Yusa K, Garnett MJ (2019) Minimal genome-wide human CRISPR-Cas9 library. bioRxiv