This study was conducted to develop a off-target-prediction pipeline and study biological importance of features using machine learning approach and statistical analyses. Multiple classification algorithms were trained on two different datasets (AsCpf1 and LbCpf1) and their performances was compared. Best performing models was used in the off-target prediction pipelines.
Statisitical significance of each position-specific nucleotides were estimated to understand their biological importance in the classification of off-targets. Other mismatch associated features were also explored and their importances were studied using statistical analysis. Machine learning approach was also used to analyze the importance of features. All the neccessary analysis codes are given in Jupyter Notebooks.
Off-target prediction pipeline is divided into two parts which are to calculate the target efficiency and predict potential off-target sites in Human genome. The pipeline is developed for the prediction of potential off-target sites of a gRNA in Human genome and the prediction of gRNA efficiency to ensure the high gRNA efficiency in experimental conditions. All the scripts of off-target prediction pipeline are written and implemented in python. For the implementation of off-target prediction pipeline in user system requires various python packages and tools which are given below.
To implement the off-target prediction pipeline in your linux-based system:
Reqirements:
• FASTA36
• ViennaRNA package
• SeqIO, SeqUtils, Seq modules of Bio package
• Biopython
• Python packages: Pandas, Numpy, sci-kit learn, joblib
• Human reference genome (Hg37)
Execution
• For Potential off-target site prediction:
cd \LocationOfOff-targetPredictionPipeline\off-target_prediction_pipeline
Python3 Fasta36_SeqSearch.py input_filename.fasta Reference_genome.fna Output_file.csv
• For target efficiency prediction:
>> cd \LocationOfOff-targetPredictionPipeline\off-target_prediction_pipeline\
For the prediction of target efficiency for AsCpf1 species
>> Python3 AsCpf1_main.py input_filename.fasta Reference_genome.fna Output_file.csv
Or For the prediction of target efficiency for LbCpf1 species
>> Python3 LbCpf1_main.py input_filename.fasta Reference_genome.fna Output_file.csv
Refer usage file for detail implementation guide for the off-target prediction pipeline
In case of any queries and suggestions, please submit a query or an issue.
(Suggestions are always welcome! 😃 )