Fig. 1.

The sequence information model uses 1D Conv and Bi-LSTM layers. First, the membrane protein sequences are encoded into a matrix of size 1500 ×20. Then, these matrixes are fed to the 1D Conv layer (filters: 256, kernel size: 15, strides: 10, padding: same, activation: relu). Next, two Bi-LSTM layers (units: 128, dropout: 0.5 and the remainder as default settings) are used to identify features separated by large gaps. Last, the output from the Bi-LSTM layer is passed through a final dense layer that uses a softmax function to obtain the probability value belonging to each membrane protein type