inDelphi-model

Dependencies

• python 2.7+ or 3.4+
• pandas 0.23.4
• scikit-learn 0.18.1 or 0.20.0
• scipy 1.1.0
• numpy 1.15.3

The online interactive web app version of inDelphi uses the scikit-learn v0.18.1 models.

Installation

Clone this github repository, then set up your environment to import the inDelphi.py script in however is most convenient for you. In python, for instance, you may use the following at the top of your script to import inDelphi.

import sys
sys.path.append('/directory/to/local/indelphi/repo/clone/')
import inDelphi

Usage

In python2.7+ or python3.4+:

import inDelphi
inDelphi.init_model(celltype = 'mESC')

Note: Supported cell types are ['mESC', 'U2OS', 'HEK293', 'HCT116', 'K562']. If your cell type of interest is not included here, we recommend using mESC if your cell type does not have known DNA repair defects and is not a cancer cell type. See www.crisprindelphi.design/guide for more details.

pred_df, stats = inDelphi.predict(seq, cutsite)

seq is a string of DNA characters. cutsite is an int that specifies the 0-index position of the cutsite, such that seq[:cutsite] and seq[cutsite:] in Python notation describe the cut products.

pred_df is a pandas dataframe containing a row for each prediction. By default, a single prediction corresponds to a single genotype for 1-bp insertions and microhomology deletions. For microhomology-less deletions, a single prediction corresponds to a single predicted frequency for the sum total frequency of a group of microhomology-less deletions. Microhomology-less deletions are grouped by deletion length.

• 1-bp insertion genotypes are uniquely identified by row['Category'] == 'ins' and row['Inserted Bases'].isin([A, C, G, T]).
• Microhomology deletions are uniquely identified by row['Category'] == 'del', 1 <= row['Length'] <= 60, and 0 <= row['Genotype position'] <= row['Length'].
• Microhomology-less deletions are uniquely identified by row['Category'] == 'del', 1 <= row['Length'] <= 60, and row['Genotype position'] == 'e'.
• The column 'Predicted frequency' sums to 100.0.

stats is a dict with the following keys. For further details, refer to www.crisprindelphi.design/guide#batch4.

• Phi (Note: natural log of phi refers to microhomology strength)
• Precision
• 1-bp ins frequency
• MH del frequency
• MHless del frequency
• Frameshift frequency
• Frame +0 frequency
• Frame +1 frequency
• Frame +2 frequency
• Highest outcome frequency
• Highest del frequency
• Highest ins frequency
• Expected indel length
• Reference sequence
• Cutsite
• gRNA
• gRNA orientation
• Cas9 type
• Celltype

Example usage

import inDelphi
inDelphi.init_model(celltype = 'mESC')

left_seq = 'AGAATCGCCCGCGGTCCATCCTTTATCAGCGGGAATTCAAGCGCACCAGCCAGAGGTGTA'
right_seq = 'CCGTGGACGTGAGAAAGAAGAAACATAATATTCGCACTAGATCCATCCCCATACCTGACC'
seq = left_seq + right_seq
cutsite = len(left_seq)

pred_df, stats = inDelphi.predict(seq, cutsite)

Additional methods

Once you have obtained pred_df, stats, additional methods are available for your convenience.

Obtaining exact genotypes

pred_df, stats = inDelphi.predict(seq, cutsite)
pred_df = inDelphi.add_genotype_column(pred_df, stats)

A new column Genotype will be created.

• If MH-less genotypes are represented with pred_df['Genotype position'] == 'e', MH-less genotypes will not have a value for Genotype since these rows represent the total predicted frequency for a group of MH-less genotypes.

Expanding microhomology-less deletion predictions into genotype resolution

Warning: Microhomology-less deletions are less consistent between experimental replicates than 1-bp insertions and microhomology deletions. inDelphi as presented in our manuscript was tested for performance only on grouped predictions for MH-less genotypes.

pred_df, stats = inDelphi.predict(seq, cutsite)
pred_df = inDelphi.add_mhless_genotypes(pred_df, stats)
# pred_df = inDelphi.add_genotype_column(pred_df, stats) 

Microhomology-less deletions will be converted into a different representation:

• Microhomology-less deletions will be uniquely identified by row['Category'] == 'del', 1 <= row['Length'] <= 60, and 0 <= row['Genotype position'] <= row['Length'].
• Microhomology-less deletions will no longer contain the value 'e' in 'Genotype position'
• The number of rows in pred_df will increase

For details on how predicted frequencies are converted, refer to https://www.crisprindelphi.design/guide#single4.

Contact

maxwshen at mit.edu

License

Limited Copyright License for Research Use by Non-Profit and Government Institutions

BY DOWNLOADING THE CODE OR USING THE SERVICE AND/OR SOFTWARE APPLICATION ACCOMPANYING THIS LICENSE, YOU ARE CONSENTING TO BE BOUND BY ALL OF THE TERMS OF THIS LICENSE

“Copyright 2018. Massachusetts Institute of Technology, The Broad Institute, Harvard University and Brigham and Women's Hospital. All Rights Reserved.”

The software is being provided as a service for research, educational, instructional and non-commercial purposes only. By generating a user account and/or submitting jobs to InDelphi you agree to the terms and conditions herein. You are an actively enrolled student, post-doctoral researcher, or faculty member at a degree-granting educational institution or US government research institution; and You will only use the InDelphi Software Application and/or Service for educational, instructional, and/or non-commercial research purposes; You understand that all results produced using the Code may only be used for non-commercial research and/or academic purposes; You understand that to obtain any right to use the Code for commercial purposes, or in the context of industrially sponsored research, You must enter into an appropriate, separate and direct license agreement with the Owners. You will not redistribute unmodified versions of the Code; You will redistribute modifications, if any, under the same terms as this license and only to non-profits and US government institutions; You must credit the authors of the Code: David K. Gifford, Jonathan Yee-Ting Hsu and Max Walt Shen and cite Predictable and precise template-free editing of pathogenic mutations by CRISPR-Cas9 nuclease", Nature, 2018, doi:10.1038/s41586-018-0686-x; and You understand that neither the names of the Owners nor the names of the authors may be used to endorse or promote products derived from this software without specific prior written permission.