Learning Proteome Domain Folding Using LSTMs in an Empirical Kernel Space. Issue 15 (15th August 2022)

• Record Type: Journal Article
• Authors: Kuang, Da; Issakova, Dina; Kim, Junhyong
• Is Part Of: Journal of molecular biology. Volume 434:Issue 15(2022)
• Publication Date: 2022-08-15
• Subjects: protein empirical structure space -- long short-term memory networks -- proteins fold -- SCOPe; Molecular biology -- Periodicals; Biology -- Periodicals; Biochemistry -- Periodicals; Bacteriology -- Periodicals; Molecular Biology -- Periodicals; Biochemistry -- Periodicals; Biologie moléculaire -- Périodiques; Biologie -- Périodiques; Biochimie -- Périodiques; Moleculaire biologie; Biochemistry; Biology; Molecular biology; Periodicals; 572.805
• DOI: 10.1016/j.jmb.2022.167686
• Languages: English
• ISSNs: 0022-2836

Abstract

Graphical abstract: Highlights: Protein fold prediction in feature space using threading scores to solved structures. Modularization by segmenting query proteins and applying recurrent LSTM network. De novo prediction of non-human species proteins from training on human sequences. Require 14% of previous studies' training data to achieve near cutting-edge accuracy. Abstract: The recognition of protein structural folds is the starting point for protein function inference and for many structural prediction tools. We previously introduced the idea of using empirical comparisons to create a data-augmented feature space called PESS (Protein Empirical Structure Space) 1 as a novel approach for protein structure prediction. Here, we extend the previous approach by generating the PESS feature space over fixed-length subsequences of query peptides, and applying a sequential neural network model, with one long short-term memory cell layer followed by a fully connected layer. Using this approach, we show that only a small group of domains as a training set is needed to achieve near state-of-the-art accuracy on fold recognition. Our method improves on the previous approach by reducing the training set required and improving the model's ability to generalize across species, which will help fold prediction for newly discovered proteins.