In this project, deep learning in the form of neural networks will be utilised to predict the EC class of input enzymes at a deep hierarchal level. Long Short-Term Memory (LSTM) and Gated Recurrent Units (GRU) binary classification models were constructed to undertake this task. These networks were chosen for the capacity to learn long term and prowess at processing sequence data.
Two enzyme datasets consisting of maximum sequence lengths of 500 and 1,000 respectively, were successfully run-on LSTM and GRU neural network models giving a total of four experiments.
However, the neural networks did not perform to the required level for classifying enzyme sequences with accuracies of just over 50 % achieved for all four models.