Redundancy-weighting the PDB for detailed secondary structure prediction using deep-learning models

Abstract

Motivation The Protein Data Bank (PDB), the ultimate source for data in structural biology, is inherently imbalanced. To alleviate biases, virtually all structural biology studies use nonredundant (NR) subsets of the PDB, which include only a fraction of the available data. An alternative approach, dubbed redundancy-weighting (RW), down-weights redundant entries rather than discarding them. This approach may be particularly helpful for machine-learning (ML) methods that use the PDB as their source for data. Methods for secondary structure prediction (SSP) have greatly improved over the years with recent studies achieving above 70% accuracy for eight-class (DSSP) prediction. As these methods typically incorporate ML techniques, training on RW datasets might improve accuracy, as well as pave the way toward larger and more informative secondary structure classes. Results This study compares the SSP performances of deep-learning models trained on either RW or NR datasets. We show that training on RW sets consistently results in better prediction of 3- (HCE), 8- (DSSP) and 13-class (STR2) secondary structures. Availability and implementation The ML models, the datasets used for their derivation and testing, and a stand-alone SSP program for DSSP and STR2 predictions, are freely available under LGPL license in http://meshi1.cs.bgu.ac.il/rw. Supplementary information Supplementary data are available at Bioinformatics online.

TopEC: Improved classification of enzyme function by a localized 3D protein descriptor and 3D Graph Neural Networks

“…For example, for translocases (F-score: 0), we only have 228 experimental structures for training, validation, and testing. Furthermore, while experimental evidence is regarded as the ground truth when creating prediction models, the PDB has known issues such as redundancy 44 or enzyme-inhibitor complexes marked as enzyme-substrate complexes 45–47 . For serine endopeptidase La (EC: 3.4.21.53), none of the eight test PDBs are correctly predicted.…”

Deep learning for protein secondary structure prediction: Pre and post-AlphaFold

“…Other studies proposed novel architectures as well as input features, for example [36] , [56] . Besides the model’s architecture and features, several studies also proposed new training and test datasets that eliminate data redundancy such that the models trained using these datasets can achieve a good performance [13] , [57] , [58] .…”

Estimation of Model Accuracy By A Unique Set of Features And Tree-Based Regressor

“…• meshinr_dssp8Compatibility and meshirw_dssp8Compatibility_Weighted -Two slightly different measures of the agreement between the secondary structure (8 states) of the decoy's residues (as measured by DSSP 55 ) and their predicted secondary structure 49 .…”

“…These terms include pair-wise atomic potentials 30,42,43 , torsion angles 44 , hydrogen bonds, and hydrogen bond patterns 45 , solvation terms, "meta" energy terms that consider the distribution of other terms within the protein atoms, an extended radius of gyration that takes into account different classes of amino acids (polar vs. non-polar, secondary structure elements vs. coil region, etc. ), and compatibility of the decoys with solvent exposure prediction 46 , and with 3-, 8-and 13-classes secondary structure predictions [47][48][49] .…”