Prediction of Transmembrane Helix Using a Compound Neural Network

In this paper, a novel method was proposed for the prediction of transmembrane helices - one of the key issues in the structure prediction of membrane protein. Concretely, the membrane protein sequence was mapped into propensity factor sequence, which then yeilds J detail sequences through J levels of discrete wavelet transform (DWT). Lastly, the original sequence and a subset of the J detail sequences were assembled and put into ScaleNet, a compound neural network which outputs the transmembrane helices sequence. The experiments show that, our method outperforms the DWT-peak-extension method, one of the perfect methods.