EdaFoldAA

EdaFoldAA predicts de novo protein structures by fragment assembly guided by an Estimation of Distribution Algorithm (EDA) to improve conformational sampling and all-atom model refinement.


Key Features:

  • Fragment Assembly Approach: Builds initial coarse-grained models from pools of candidate fragments sourced from known protein structures and refines them to all-atom models.
  • Stochastic Sampling Refinement: Uses stochastic sampling to explore complex energy landscapes, generating numerous independent models due to randomness in the search.
  • Iterative Distribution Estimation (EDA): Iteratively refines a distribution over fragments based on low-energy all-atom models to bias subsequent fragment selection toward native-like regions.
  • All-Atom Energy Function: Applies an all-atom energy function to evaluate and refine models at atomic resolution.
  • Benchmark Performance: Demonstrated ability to reach lower energy levels and produce a higher percentage of near-native models in blind benchmark selections (reported relative to protocols such as AbInitioRelax).

Scientific Applications:

  • De novo protein structure prediction: Predicts tertiary structures for proteins lacking homologous experimental structures.
  • Structural biology: Supports modeling of protein conformations for studies of fold and stability when experimental data are limited.
  • Drug discovery: Provides high-resolution models useful as starting points for structure-based ligand design and target characterization.
  • Enzyme design: Supplies atomic-detail models to guide mutation design and catalytic site engineering.
  • Protein function and interaction analysis: Generates structures that aid interpretation of molecular function and potential interaction interfaces.

Methodology:

Build initial models from a pool of candidate fragments, refine models via stochastic sampling to explore the energy landscape, iteratively estimate and refine fragment distributions using an EDA based on low-energy all-atom models, and evaluate/refine models with an all-atom energy function.

Topics

Details

Tool Type:
command-line tool
Operating Systems:
Linux
Added:
12/18/2017
Last Updated:
11/25/2024

Operations

Publications

Simoncini D, Zhang KYJ. Efficient Sampling in Fragment-Based Protein Structure Prediction Using an Estimation of Distribution Algorithm. PLoS ONE. 2013;8(7):e68954. doi:10.1371/journal.pone.0068954. PMID:23935913. PMCID:PMC3723781.

Links