Dataset Information

SAINT-Angle: self-attention augmented inception-inside-inception network and transfer learning improve protein backbone torsion angle prediction.

ABSTRACT:

Motivation

Protein structure provides insight into how proteins interact with one another as well as their functions in living organisms. Protein backbone torsion angles ( ϕ and ψ ) prediction is a key sub-problem in predicting protein structures. However, reliable determination of backbone torsion angles using conventional experimental methods is slow and expensive. Therefore, considerable effort is being put into developing computational methods for predicting backbone angles.

Results

We present SAINT-Angle, a highly accurate method for predicting protein backbone torsion angles using a self-attention-based deep learning network called SAINT, which was previously developed for the protein secondary structure prediction. We extended and improved the existing SAINT architecture as well as used transfer learning to predict backbone angles. We compared the performance of SAINT-Angle with the state-of-the-art methods through an extensive evaluation study on a collection of benchmark datasets, namely, TEST2016, TEST2018, TEST2020-HQ, CAMEO and CASP. The experimental results suggest that our proposed self-attention-based network, together with transfer learning, has achieved notable improvements over the best alternate methods.

Availability and implementation

SAINT-Angle is freely available as an open-source project at https://github.com/bayzidlab/SAINT-Angle.

Supplementary information

Supplementary data are available at Bioinformatics Advances online.

SUBMITTER: Hasan AKMM

PROVIDER: S-EPMC10115468 | biostudies-literature | 2023

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

SAINT-Angle: self-attention augmented inception-inside-inception network and transfer learning improve protein backbone torsion angle prediction.

Hasan A K M Mehedi AKMM Ahmed Ajmain Yasar AY Mahbub Sazan S Rahman M Saifur MS Bayzid Md Shamsuzzoha MS

Bioinformatics advances 20230405 1

<h4>Motivation</h4>Protein structure provides insight into how proteins interact with one another as well as their functions in living organisms. Protein backbone torsion angles ( ϕ and ψ ) prediction is a key sub-problem in predicting protein structures. However, reliable determination of backbone torsion angles using conventional experimental methods is slow and expensive. Therefore, considerable effort is being put into developing computational methods for predicting backbone angles.<h4>Resul ...[more]

PMID: 37092035

Similar Datasets

Project description:This groundbreaking study derives and tests several new dihedral torsion model potentials for constructing classical forcefields for atomistic simulations of materials. (1) The new angle-damped dihedral torsion (ADDT) model potential is preferred when neither contained equilibrium bond angle is linear (i.e., (θ eq ABC and θ eq BCD) ≠ 180°), at least one of the contained equilibrium bond angles is ≥ 130° (i.e., (θ eq ABC or θ eq BCD) ≥ 130°), and the dihedral torsion potential contains some odd-function contributions (i.e., U[ϕ] ≠ U[-ϕ]). (2) The new angle-damped cosine only (ADCO) model potential is preferred when neither contained equilibrium bond angle is linear (i.e., (θ eq ABC and θ eq BCD) ≠180°), at least one of the contained equilibrium bond angles is ≥ 130° (i.e., (θ eq ABC or θ eq BCD) ≥ 130°), and the dihedral torsion potential contains no odd-function contributions (i.e., U[ϕ] = U[-ϕ]). (3) The new constant amplitude dihedral torsion (CADT) model potential is preferred when neither contained equilibrium bond angle is linear (i.e., (θ eq ABC and θ eq BCD) ≠ 180°), both contained equilibrium bond angles are <130° (i.e., (θ eq ABC and θ eq BCD) < 130°), and the dihedral torsion potential contains some odd-function contributions (i.e., U[ϕ] ≠ U[-ϕ]). (4) The constant amplitude cosine only (CACO) model potential is preferred when neither contained equilibrium bond angle is linear (i.e., (θ eq ABC and θ eq BCD) ≠180°), both contained equilibrium bond angles are <130° (i.e., (θ eq ABC and θ eq BCD) <130°), and the dihedral torsion potential contains no odd-function contributions (i.e., U[ϕ] = U[-ϕ]). (5) The new angle-damped linear dihedral (ADLD) model potential is preferred when at least one contained equilibrium bond angle is linear (i.e., (θ eq ABC or θ eq BCD) = 180°). Most importantly, this article derives combined angle-dihedral coordinate branch equivalency conditions and angle-damping factors that ensure the angle-damped torsion model potentials (e.g., ADDT, ADCO, and ADLD) are mathematically consistent and continuously differentiable even as at least one contained bond angle approaches linearity (i.e., as (θ ABC or θ BCD) → 180°). This article introduces the torsion offset potential (TOP). I show the TOP gives rise in some materials to the unusual physical phenomenon of slip torsion. For various molecules, extensive quantitative comparisons to high-level quantum chemistry calculations (e.g., CCSD) and experimental vibrational frequencies showed these new dihedral torsion model potentials perform superbly.

Dataset Information

SAINT-Angle: self-attention augmented inception-inside-inception network and transfer learning improve protein backbone torsion angle prediction.

Motivation

Results

Availability and implementation

Supplementary information

Publications

SAINT-Angle: self-attention augmented inception-inside-inception network and transfer learning improve protein backbone torsion angle prediction.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets