![]() Some studies also emphasized on input feature selection. Similar to SPIDER3, NetSurfP-2.0 ( Klausen et al., 2019) used the bidirectional recurrent neural network to capture the long-range interactions. MUFOLD ( Fang et al., 2018b) used deep residual inception models ( Szegedy et al., 2017) to measure the short-range and long-range interactions among different amino acid residues. SPIDER2 ( Heffernan et al., 2015) used iterative neural network to predict the backbone torsion angles, while SPIDER3 ( Heffernan et al., 2017) leveraged the bidirectional recurrent neural network (BiRNN) ( Schuster and Paliwal, 1997) to capture the long-range interactions among amino acid residues in a protein molecule. Several deep learning-based techniques have recently been developed that can predict backbone torsion angles with a reasonable accuracy. ![]() Real-SPINE ( Dor and Zhou, 2007) leveraged an integrated system of neural networks to predict the real values of dihedral angles. Given the growing availability of protein databases and rapid advances in machine learning (ML) methods (especially, the deep learning techniques), application of ML techniques to leverage the available data in accurate prediction of backbone angles has gained significant attention.Įarlier ML-based methods used neural network ( Wu and Zhang, 2008), support vector machine (SVM) ( Wu and Zhang, 2008) and hidden Markov model (HMM) ( Bystroff et al., 2000 Karchin et al., 2003) to predict discrete states of torsion angles ϕ and ψ. As a result, accurate prediction of torsion angles can significantly advance our understanding of the 3D structures of proteins. Therefore, protein structure prediction is often divided into smaller and more doable sub-problems ( Heffernan et al., 2017) such as backbone torsion angles prediction. The backbone torsion angles play a critical role in protein structure prediction and investigating protein folding ( Adhikari et al., 2012 Gao et al., 2018 Tian et al., 2020). Therefore, developing efficient computational approaches for determining protein structures has been gaining increasing attention from the scientific community ( AlQuraishi, 2019 Greener et al., 2019 Senior et al., 2020 Xu, 2019 Xu et al., 2020). However, the experimental determination of protein structures using X-ray crystallography, cryogenic electron microscopy (cryo-EM) and nuclear magnetic resonance spectroscopy is costly and time- and labour-intensive ( Jiang et al., 2017). Proteins are responsible for various functions in cells and their functions are usually determined by their 3D structures.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |