Jutan Das

CRISPR-Cas9 system is one of the most significance genome editing techniques in the recent time period because of its higher potentiality to modify the specific target genes and region of the genome which are complementary of the designed guide RNA (or sgRNA). Based on the target sequence different sgRNA was design for accurately manipulating the desired genomic sites. CRISPR-Cas9 still now suffering from the off-target effect. Here, I developed three machine learning based techniques (i.e. Artificial Neural Network, Support Vector Machine and Random Forest) for estimation of the CRISPR-Cas9 cleavage sites to be cleaved by a given sgRNA. All of these machine learning model are developed based on the plant dataset which are overlooked in previous studies. The models were train and tested on the collected on-target and off-target dataset of different plant species. For ANNs I developed total six models (ANN1-Logistic, ANN1-Tanh, ANN1-ReLU, ANN2-Logistic, ANN2-Tanh, and ANN-ReLU) among all of these model ANN1- ReLU model give the best result from other ANN based developed models. Here, I developed total four SVM model (SVM-Linear, SVM-Polynomial, SVM-Gaussian and SVM-Sigmoid), from all the model SVM-Linear model performance better compare to other three developed model. I demonstrate that random forest model attains the best performance on the plant dataset among other developed models, its produce an average classification area under the ROC curve (AUC) is 99.0%. Also I display that the prediction by my developed models are more precise compare to other available methodologies (CRISTA). Additional analyses are led to explore the fundamental reasons from different perceptions.