SPROUT is a machine learning algorithm to predict the repair outcome of a CRISPR-CAS9 knockout experiment. SPROUT accepts the DNA sequence of the guide as input as well as other genomic factors. It then predicts various statistics of the repair outcome, including: the fraction of mutant reads with an insertion/deletion, fraction of total reads with insertion/deletion, average insertion length given an insertion, average deletion length given a deletion, diversity, most likely inserted base pair and finally the edit (mutation) efficiency of the CRISPR outcome. In this repository we provide the codes neccessary to run the SPROUT algorithm as well as the set of trained models for users who want to run SPROUT in the batch mode. We have also provided an easy-to-use SPROUT website with graphical interfance for general users with limited access to computing resources.
Details of SPROUT is provided in the manuscript "Systematic Characterization of Genome Editing in Primary T cells Reveals Proximal Genomic Insertions and Enables Machine Learning Prediction of CRISPR-Cas9 DNA Repair Outcomes".
This repository provides the SPROUT package which includes the scripts required to predict the DNA repair outcome as well as a complete set of trained SPROUT models. The trained models are stored in the model folder of the repository. The main script of the package is the SPROUT_predict script which loads the pretrained models, asks for input query depending on the mode selected, and outputs the predicted statistics of the repair outcome.
To use the light-weight prediction tool run the SPROUT_prediction script and follow the instruction. The only software requirement is python 2.7.X. or later. SPROUT pipeline conveniently works in three modes depending on what input format is available to the user. Here we detail each operating mode:
Mode (1) The first mode takes the DNA sequence of the guide and PAM sequence (23 base pair sequence) as the input. No other input is required for this operational mode. This is ideal for nucleotide-only guide design purposes. The code asks for one of the three input operating modes.
Which input format do you prefer?
(1) sgRNA sequence only
(2) sgRNA sequence + genomic features (chromatin, etc.)
(3) location on the genome and cell type
For the basic mode input the digit 1.
Selected option:
1
Now the code asks for the sgRNA sequence followed by the PAM sequence.
Input the sgRNA sequence followed by the PAM sequence:
TATGCATGCATCGACGATCGGGG
The software output SPROUT's prediction results as detailed in the manuscript.
Here are the repair outcomes that SPROUT predicts for this guide:
Fraction of total reads with insertion 19 %
Insertion to deletion ratio 27 %
Average insertion length 1.1 bps
Average deletion length 12.0 bps
Diversity 2.25 (Low)
Most likely inserted base pair A
Mode (2) The second mode accepts the input guide sequence as well as the 33 genomic features as listed in the main text. The input format is a string with 33 features, separated by comma. The ordering of features are as follows:
GC, CpG, priPhCons, mamPhCons, verPhCons, priPhyloP, mamPhyloP, verPhyloP, GerpN, GerpS, GerpRS, bStatistic, fitCons, cHmmTssA, cHmmTssAFlnk, cHmmTxFlnk, cHmmTx, cHmmTxWk, cHmmEnhG, cHmmEnh, cHmmZnfRpts, cHmmHet, cHmmTssBiv ,cHmmBivFlnk, cHmmEnhBiv, cHmmReprPC, cHmmReprPCWk, cHmmQuies, EncExp, EncH3K27Ac, EncH3K4Me1, EncH3K4Me3, EncNucleo.
Details of the features are described in Supplementary Fig. 5 as adapted from the ENCODE project detailed "here". This operation mode of SRPOUT is ideal when both the guide sequence and the genomic factors are known, measured or extracted a priori. As demonstrated in the manuscript the chromatin facots do not effect
Selected option:
2
Input the sgRNA sequence followed by the PAM sequence:
TATGCATGCATATATATATAGGG
Input the genomic factors separated by ',':
0,0.5,0,0,0.5,0,0,0.5,0,0,0.5,0,0,0.5,0,0,0.5,0,0,0.5,0,0,0.5,0,0,0.5,0,0,0.5,0,0,0.5,0
And the software output the predicted summary statistics.
Here are the repair outcomes that SPROUT predicts for this guide:
Fraction of total reads with insertion 22 %
Insertion to deletion ratio 58 %
Average insertion length 0.8 bps
Average deletion length 11.5 bps
Diversity 2.92 (Low)
Most likely inserted base pair T
Edit efficiency 55 %
Mode (3) (Currently under construction) The third operation mode asks the coordinate of the cut site and the cell type as the input. The SPROUT software extracts the guide sequence as well as the PAM from the human refrence genome (hg38.fa). The input coordinate should be the index of the starting nucleotide of the guide sequence (from the 3’ end). Once the guide is extracted, the softwares does a few tests to validate the guide and PAM sequences. Then, the software extracts the genomic factors given the cell type from a server. Finally, the SPROUT software predicts the out given the input sequence and chromatin facotrs.
Selected option:
3
What is the target cell type?
T-cells
Which chromosome the cut site locates?
1
What location the cut site is in that chromosome?
1234567
This is the selected guide sequence:
CGTTGAGTTCGAGCTCCGATGGG
Here are the repair outcomes that SPROUT predicts for this guide:
Fraction of total reads with insertion 17 %
Insertion to deletion ratio 50 %
Average insertion length 1.0 bps
Average deletion length 8.3 bps
Diversity 2.49 (Low)
Most likely inserted base pair G