This project is intended for study of implementing deep learning to a binary classification of sgRNA off-target prediction. Five different models were compared in this study to get the best performance in classification prediction for off-target sgRNA. All models implement Word2Vec embedding to get better feature vector representation compared to the traditional one-hot encoding method. The five models were constructed by different RNN models: biLSTM, LSTM, GRU, biGRU, and without RNN Layer (NoRNN). Each model was trained with a learning rate default of Adam optimizer 0.001 and tested with two different datasets: HEK293T dataset and K562 Dataset.
Paper has been published on IEEE and presented on ICCoSITE 2023 Conference
Check this repository for the full project
The work of this project aims to study the application of machine learning in the field of bioinformatics, especially Drug Discovery. This project was created by following a tutorial from Chanin Nantasenamat. Then this project continued for the Natural Language Processing coursework at Binus University by fine-tuning the dataset with the Transformers model (ChemBERTa).
Paper has been published on IEEE and presented on ICITISEE 2023 Conference
Check this repository for the full project