Improving protein disorder prediction by deep bidirectional long short‐term memory recurrent neural networks

By Jack Hanson, Yuedong Yang, Kuldip K. Paliwal, Yaoqi Zhou. 2016

250 citations recorded by the source corpus.

Motivation: Capturing long‐range interactions between structural but not sequence neighbors of proteins is a long‐standing challenging problem in bioinformatics. Recently, long short‐term memory (LSTM) networks have significantly improved the accuracy of speech and image classification problems by remembering useful past information in long sequential events. Here, we have implemented deep bidirectional LSTM recurrent neural networks in the problem of protein intrinsic disorder prediction. Results: The new method, named SPOT‐Disorder, has steadily improved over a similar method using a traditional, window‐based neural network (SPINE‐D) in all datasets tested without separate training on short and long disordered regions. Independent tests on four other datasets including the datasets from critical assessment of structure prediction (CASP) techniques and >10 000 annotated proteins from MobiDB, confirmed SPOT‐Disorder as one of the best methods in disorder prediction. Moreover, initial studies indicate that the method is more accurate in predicting functional sites in disordered regions. These results highlight the usefulness combining LSTM with deep bidirectional recurrent neural networks in capturing non‐local, long‐range interactions for bioinformatics applications. Availability and Implementation: SPOT‐disorder is available as a web server and as a standalone program at: http://sparks‐lab.org/server/SPOT‐disorder/index.php. Contact: j.hanson@griffith.edu.au or yuedong.yang@griffith.edu.au or yaoqi.zhou@griffith.edu.au Supplementary information: Supplementary data is available at Bioinformatics online.

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

SPOT-Disorder2: Improved Protein Intrinsic Disorder Prediction by Ensembled Deep Learning (2019, methodology): “SPOT-Disorder [19], previously introduced by us, employed long short-term memory (LSTM) cells [22] in a bidirectional recurrent neural network (BRNN) [23] for protein disorder prediction.”
NetSurfP-2.0: improved prediction of protein structural features by integrated deep learning (2018, methodology): “In fact, when assessed on various test sets with less than 25% sequence identity to any protein used in the training, its accuracy was consistently on par with or better than that of other stateof-the-art tools (3,4,17,21,22).”
AlphaFold2: A Role for Disordered Protein/Region Prediction? (2022, methodology): “We selected the top 10 performing on the DisProt and DisProt-PDB giving a combined non-redundant set of 11 (fIDPnn [34], SPOT-Disorder2 [35], RawMSA [36], fIDPlr [34], PreDisorder [37], AUCpreD [38], SPOT-Disorder1 [39], SPOT-Disorder-Single (SPOT-Disorder-S) [40] , DisoMine [41], AUCpreD-np [38] an”
A Survey of Deep Learning for Scientific Discovery (2020, methodology): “…popular benchmark in natural language, and also for capturing sequence dependencies in scientiﬁc applications, such as protein function prediction [77, 80], and using the hidden representations as part of a larger pipeline for protein structure prediction in AlphaFold [205] (with opensourced code…”
Protein remote homology detection based on bidirectional long short-term memory (2017, background): “…of bioinformatics [39], such as the estimation of protein model quality [40], protein structure prediction [41 – 43], protein disorder prediction [44], etc. Recurrent Neural Network (RNN) is one of the most successful deep learning techniques, which is designed to utilize sequential information…”
Sixty-five years of the long march in protein secondary structure prediction: the final stretch? (2016, background): “Indeed, the application of the LSTM network to protein disorder prediction has confirmed its improvement over regular neural networks in detecting structured and unstructured regions [160].”