Design a new Method for Prediction of DNA-binding Protein using Deep Learning
Abstract
A protein that can bind with DNA is called as DNA binding protein. It can also interact with DNA. It has an important and critical role in gene expression and transcription. Therefore, using DNA binding protein we can develop some important drugs which can be used to treat cancers and genetic diseases. So in the domain of molecular biology, it has become a challenging and essential problem for researchers to develop highly accurate and efficient methods for identifying DNA binding protein. Also, the experimental methods are more time consuming and very expensive, hence there is a need for a method based on machine learning. Here, various experiments are demonstrated and analyzed. Finally, we have proposed a method to predict DNA-binding protein using Convolutional Neural Networks. This proposed method takes a 2D PSSM (Position Specific Scoring Matrix) of a protein sequence as input with the dataset that is available in Protein Data Bank. We have attained an accuracy of 97.67% when we have performed our method on the PDB1075 dataset, and it has attained an accuracy of 89.32% on the PDB186 dataset.