Protein Secondary Structure Prediction with Long Short Term Memory Networks

By Søren Kaae Sønderby, Ole Winther. 2014

117 citations recorded by the source corpus.

Prediction of protein secondary structure from the amino acid sequence is a classical bioinformatics problem. Common methods use feed forward neural networks or SVMs combined with a sliding window, as these models does not naturally handle sequential data. Recurrent neural networks are an generalization of the feed forward neural network that naturally handle sequential data. We use a bidirectional recurrent neural network with long short term memory cells for prediction of secondary structure and evaluate using the CB513 dataset. On the secondary structure 8-class problem we report better performance (0.674) than state of the art (0.664). Our model includes feed forward networks between the long short term memory cells, a path that can be further explored.

Subfields: Molecular Biology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Molecular Biology

Citations in the corpus

Deep learning for computational biology (2016, neutral): “…et al , 2014; Sutskever et al , 2014; Lipton, 2015; Xiong et al , 2016), and most recently, in computational biology (Eickholt & Cheng, 2013; Dahl et al , 2014; Leung et al , 2014; Sønderby & Winther, 2014; Alipanahi et al , 2015; Wang et al , 2015; Zhou & Troyanskaya, 2015; Kelley et al , 2016).”
Machine-learning-guided directed evolution for protein engineering (2018, background): “…acid binding 46–48 , protein–major histocompatibility complex binding 49 , binding-site prediction 50 , protein–ligand binding 51,52 , solubility 53 , thermostability 54,55 , subcellular localization 56 , secondary structure 57 , functional class 58–60 , and even three-dimensional structure 61 .”
Machine learning in protein engineering (2018, background): “…(Khurana et al., 2018), thermostability (Dehouck et al., 2009; Giollo et al., 2014), subcellular localization (Almagro Armenteros et al., 2017), secondary structure (Sønderby & Winther, 2014), functional class (Szalkai & Grolmusz, 2018; Cao et al., 2017), and even 3D structure (Hopf et al., 2012).”
LSTM: A Search Space Odyssey (2015, background): “2582924 structure prediction [11], analysis of audio [12], and video”
Exploring evolution-based & -free protein language models as protein function predictors (2022, background): “Secondary structure (SS) prediction [35] is a easier task than predicting the 2D contact map [42] and 3D structure [20].”
Deep learning in bioinformatics (2016, methodology): “Protein structure [88-90] Brain decoding [168] Gene expression regulation [105-107] Anomaly classification [177, 178] Protein classification [109, 110]”
Deep Learning and Its Applications in Biomedicine (2018, background): “Søn-derby et al. used a bidirectional RNN (BRNN) with long short-term memory cells to improve the prediction of secondary structure, with better accuracy (0.671) than that using state of the art (0.664) [178].”
An introduction to deep learning on biological sequence data: examples and solutions (2017, methodology): “The architecture of our neural network to predict secondary is shown in Figure 3B and is a combination of convolutional and LSTM neural networks (Sønderby and Winther, 2014).”
Next-Step Conditioned Deep Convolutional Neural Networks Improve Protein Secondary Structure Prediction (2017, background): “…of neural networks for secondary structure prediction was pi-oneered by Qian and Sejnowski in 1988, and subsequent attempts using both recurrent (Sønderby & Winther, 2014; Li & Yu, 2016) and convolutional (Li & Yu, 2016; Zhou & Troyanskaya, 2014; Lin, Lanchantin, & Qi, 2016; Wang et al ., 2016)…”