DeeperBind is an extension of DeepBind which adds a layer of Deep LSTM to model positional information. DeeperBind is deisgned to predict protein-DNA binding affinity from high-throughput assays that measure the binding affinity.
##Architecture
We used Deep Convolutional Neural Network to chracterize multiple motifs and the long-short term memory networks to capture temporal (i.e. positional) features on the probe sequences. DeeperBind has been successfully tested on Proteim Binding Microarray (PBM) data but can readily be extended to other data such as chip-seq. For more information, please refer to this article.
##About the source code The code is written with [Lua/Torch] (http://torch.ch/). ##References [1] 1. Hassanzadeh, Hamid Reza, and May D. Wang. "DeeperBind: Enhancing Prediction of Sequence Specificities of DNA Binding Proteins." Bioinformatics and Biomedicine (BIBM), 2016 IEEE International Conference on. IEEE, 2016.