SPINE X: Improving protein secondary structure prediction by multistep learning coupled with prediction of solvent accessible surface area and backbone torsion angles

Abstract

Accurate prediction of protein secondary structure is essential for accurate sequence alignment, three-dimensional structure modeling, and function prediction. The accuracy of ab initio secondary structure prediction from sequence, however, has only increased from around 77% to 80% over the past decade. Here, we developed a multi-step neural-network algorithm by coupling secondary structure prediction with prediction of solvent accessibility and backbone torsion angles in an iterative manner. Our method called SPINE X was applied to a dataset of 2640 proteins (25% sequence identity cutoff) previously built for the first version of SPINE and achieved a 82.0% accuracy based on ten-fold cross validation (Q3). Surpassing 81% accuracy by SPINE X is further confirmed by employing an independently built test dataset of 1833 protein chains, a recently built dataset of 1975 proteins and 117 CASP 9 targets (Critical Assessment of Structure Prediction techniques) with an accuracy of 81.3%, 82.3% and 81.8%, respectively. The prediction accuracy is further improved to 83.8% for the dataset of 2640 proteins if the DSSP assignment employed above is replaced by a more consistent consensus secondary structure assignment method. Comparison to the popular PSIPRED and CASP-winning structure-prediction techniques is made. SPINE X predicts number of helices and sheets correctly for 21.0% of 1833 proteins, compared to 17.6% by PSIPRED. It further shows that SPINE X consistently makes more accurate prediction in helical residues (6%) without over prediction while PSIPRED makes more accurate prediction in coil residues (3–5%) and over predicts them by 7%. SPINE X Server and its training/test datasets are available at http://sparks.informatics.iupui.edu/

Low-Homology Protein Structural Class Prediction from Secondary Structure Based on Visibility and Horizontal Visibility Network

“…Yu et al (2017) use Chous pseudo amino acid composition and wavelet denoising to prediction structural class. From 2014 to now, several papers (Dehzangi et al, 2014;Wang et al, 2014;Jones, 1999;Faraggi et al, 2012) show that the protein secondary structure is significanc to predict protein structural classes. Firstly the features are extracted, secondly all kinds of algorithms can be used to implement the classification prediction, such as Fisher's linear discriminant algorithm (Yang et al, 2009), Support Vector Machine (SVM) (Cai et al, 2003) and so on.…”

A Segmentation-Based Method to Extract Structural and Evolutionary Features for Protein Fold Recognition

“…We also use predicted secondary structure using SPINE-X which was recently proposed by [46] and attained better results (especially for the coded area) than PSIPRED on predicting protein secondary structure [47]. Given a protein sequence, it returns an L Â 3 matrix (which will be referred to as SPINE-M for the rest of this study) consisting of the normalized probability of contribution of a given amino acid based on its position along the protein sequence to build one of the three secondary structure elements namely, a-helix, b-strands, and coils.…”

“…In this study, we will refer to this sequence as the structural consensus sequence. It is expected that predicted secondary structure using SPINE-X provides significant structural information for the PFR similar to or even better than PSIPRED due to its better performance [17], [23], [30], [46].…”

“…As it is highlighted in [49], the most sensitive methods for fold recognition use sequence profiles to represent both the query and the data base proteins. The robustness and sensitivity of PSSM and SPINE-X for feature extraction have been addressed in [23], [46], [49]. In continuation, the global and local features extracted in this study will be explained in detail.…”

Improved protein disorder predictor by smoothing output

“…The residue level information includes: (a) single valued amino acid type (all the necessary information for the correct folding of a protein is encoded in its amino acid sequence [26]); (b) seven physicochemical properties of amino acid (different types, short or long, disordered regions in protein are found to have distinguished physicochemical properties); (c) twenty PSSM's (position specific scoring matrix) indicating the evolutionary information accumulated in each residue position of a protein sequence; (d) three predicted secondary structure (helix, strand and coil) probabilities from SPINE-X [27], one predicted accessible surface area (ASA) normalized by the ASA of an extended conformation (Ala-XAla) [28] and two predicted backbone torsion angle (phi, psi) fluctuations [29] since disordered residues are characterized by lack of stable secondary structure [30], highly exposed area and angle fluctuations; (e) one monogram and twenty bigrams computed from PSSM [31] representing the conserved evolutionary information of PSSM transformed from primary structure level to three dimensional structure level, which are normalized by the median of normal density distribution of monogram and bigram values in their logarithmic space; (f) one indicator for terminal residues (five residues from Nterminal as {−1.0, −0.8, −0.6, −0.4, −0.2}, five residue from C-terminal from {+1.0, +0.8, +0.6, +0.4, +0.2} respectively, with the rest as 0.0). Finally, before feeding the features into the classifier, neighboring residue's information is aggregated using a sliding window of 21 residues (10 residues on each residue to be predicted), resulting in 21 × 56 = 1176 features per residue.…”