/DeepBE

Primary LanguagePHP

DeepBE

Predicting base editing outcomes of 63 base editors

- If downloading models for use, please download them directly from the release note.

System Requirements

    Python   3.6.13
    Python Packages:
    numpy 1.19.5
    pandas 1.3.0

    Tensorflow and dependencies:
    Tensorflow  2.6.2
    CUDA       11.2.0
    cuDNN       8.1.0

Demo Instructions (required time, <1 min)

(1) DeepCas9variants

  • Predicting on-target efficiencies of 9 SpCas9 variants (SpCas9, VRQR variant, SpCas9-NG, SpCas9-NRRH, SpCas9-NRTH, SpCas9-NRCH, SpG, SpRY, Sc++)

  • PAM_variant feature: SpCas9 - 0, VRQR variant - 1, SpCas9-NG - 2, SpCas9-NRRH - 3, SpCas9-NRTH - 4, SpCas9-NRCH - 5, SpG - 6, SpRY - 7, Sc++ - 8

  • Input1: ./input_example.csv # List of Target Sequence(s)

    • File format:
    Target number	30 bp target sequence (4 bp + 20 bp protospacer + PAM + 3 bp), PAM_variant feature		
    GGATGACTACGCCTCTGCCTTagtAGGTCA, 2

  • Input2: ./PAM_variant_(*Cas9_variant_name)_model.h5 # Pre-trained Model Files

  • Output: ./prediction_result.xlsx

    target + PAM	feature	Prediction score
    GGATGACTACGCCTCTGCCTTagtAGGTCA	1	42.5612144470215
  • Run script:
    python ./PAM_variant_(*Cas9_variant_name).py

(2) DeepNG-BE

  • Predicting base editing outcomes of 7 base editors with SpCas9-NG

  • ABE: SpCas9-NG-ABE8e(V106W), SpCas9-NG-ABE8.17m+V106W

  • CBE: SpCas9-NG-YE1-BE4max, SpCas9-NG-SsAPOBEC3B

  • CGBE: SpCas9-NG-CGBE1, SpCas9-NG-miniCGBE1, SpCas9-NG-APOBEC-nCas9-Ung

  • Input1: ./input_example.csv # List of Target Sequence(s)

    • File format:
    Target number	30 bp target sequence (4 bp + 20 bp protospacer + PAM + 3 bp), PAM_variant feature		
    GGATGAACAACAAACTGCCTTAGTAGGTCA, 2

  • Input2: ./DeepNG-BE_(*base_editor_name)_model/ # Pre-trained Model Files

  • Output: ./prediction_result.xlsx

    target + PAM	edited output   PAM_variant   Prediction score
    GGATGAACAACAAACTGCCTTAGTAGGTCA	 GGATGAACAACAAgCTGCCTTAGTAGGTCA   2	0.00129620986990631
    GGATGAACAACAAACTGCCTTAGTAGGTCA	 GGATGAACAACAgACTGCCTTAGTAGGTCA   2	0.000881725805811584
    GGATGAACAACAAACTGCCTTAGTAGGTCA	 GGATGAACAACAggCTGCCTTAGTAGGTCA   2	0.0000012952218639839
    ...
  • Run script:
    python ./DeepNG-BE_(*base_editor_name).py

(3) DeepBE

  • Predicting base editing outcomes of 63 base editors

  • SpCas9 variants: SpCas9, VRQR variant, SpCas9-NG, SpCas9-NRRH, SpCas9-NRTH, SpCas9-NRCH, SpG, SpRY, Sc++

  • ABE: ABE8e(V106W), ABE8.17m+V106W

  • CBE: YE1-BE4max, SsAPOBEC3B

  • CGBE: CGBE1, miniCGBE1, APOBEC-nCas9-Ung

  • Input1: ./input_example.csv # List of Target Sequence(s)

    • File format:
    Target number	30 bp target sequence (4 bp + 20 bp protospacer + PAM + 3 bp), PAM_variant feature		
    GGATGAACAACAAACTGCCTTAGTAGGTCA, 5

  • Input2: ./DeepBE_(*base_editor_name)_model/ # Pre-trained Model Files

  • Output: ./prediction_result.xlsx

Output: ./prediction_result.xlsx
    target + PAM	edited output   PAM_variant   Prediction score
    GGATGAACAACAAACTGCCTTAGTAGGTCA	 GGATGAACAACAAgCTGCCTTAGTAGGTCA   2	0.00635114079341292
    GGATGAACAACAAACTGCCTTAGTAGGTCA	 GGATGAACAACAgACTGCCTTAGTAGGTCA   2	0.0043202587403357
    GGATGAACAACAAACTGCCTTAGTAGGTCA	 GGATGAACAACAggCTGCCTTAGTAGGTCA   2	6.34630623608246E-06
    ...
  • Run script:
    python ./DeepBE_(*base_editor_name).py