Related Experiment Video
Updated: Jun 23, 2026

16:41
A Protocol for Computer-Based Protein Structure and Function Prediction
Published on: November 3, 2011
Comparative benchmarking of template-based, evolutionary-diffusion, and generative language models for IsPETase
Berkay Orçun Yener1, Şurhan Göl1, Bora Kutlu1
1Department of Genetics and Bioengineering, Engineering Faculty, Alanya Alaaddin Keykubat University, Alanya 07425, Turkey.
Journal of Bioinformatics and Computational Biology
|June 22, 2026
Summary
Accurate enzyme engineering relies on precise protein models. While generative language models offer rapid exploration, homology modeling and evolutionary approaches provide superior thermodynamic accuracy for functional active site design.
Area of Science:
- Biochemistry and structural biology
- Computational biology and bioinformatics
- Enzyme engineering and protein design
Background:
- Accurate protein structure prediction is essential for rational enzyme engineering.
- High-fidelity models are required to guide modifications for desired enzyme functions.
- The performance of different prediction methods needs rigorous benchmarking.
Purpose of the Study:
- To benchmark three distinct protein structure prediction paradigms: homology modeling (SWISS-MODEL), MSA-conditioned diffusion (AlphaFold 3), and generative language modeling (ESM-3).
- To evaluate the predicted models against the experimental crystal structure of IsPETase using stereochemical validation, molecular docking, and molecular dynamics simulations.
- To assess the suitability of different prediction approaches for functional active site engineering.
Main Methods:
- Comparative analysis of SWISS-MODEL, AlphaFold 3, and ESM-3 against the experimental IsPETase crystal structure.
- Stereochemical validation to assess geometric quality of predicted models.
- Molecular docking simulations using a PET dimer analogue to evaluate active site accessibility.
- Molecular dynamics simulations to assess the stability and dynamic behavior of predicted structures.
Main Results:
- All methods reproduced the overall protein fold and preserved catalytic triad geometry.
- ESM-3 exhibited increased steric clashes and reduced hydrogen bonding compared to other methods.
- Molecular dynamics simulations showed the experimental structure and SWISS-MODEL were most stable; ESM-3 displayed greater fluctuations.
- Blind docking simulations revealed ESM-3's active site was sterically occluded and inaccessible to the substrate analogue, even after minimization.
Conclusions:
- Generative language models like ESM-3 excel at rapid scaffold exploration but lack the thermodynamic precision of established methods for active site engineering.
- Homology modeling and evolutionary approaches currently offer superior fidelity for designing functional enzyme active sites.
- The choice of protein structure prediction method is critical and depends on the specific requirements of enzyme engineering tasks.
Related Concept Videos
Evolutionary Relationships through Genome Comparisons
Genome comparison is one of the excellent ways to interpret the evolutionary relationships between organisms. The basic principle of genome comparison is that if two species share a common feature, it is likely encoded by the DNA sequence conserved between both species. The advent of genome sequencing technologies in the late 20th century enabled scientists to understand the concept of conservation of domains between species and helped them to deduce evolutionary relationships across diverse...
Conserved Binding Sites
Many proteins’ biological role depends on their interactions with their ligands, small molecules that bind to specific locations on the protein known as ligand-binding sites. Ligand-binding sites are often conserved among homologous proteins as these sites are critical for protein function.
Binding sites are often located in large pockets, and if their location on a protein’s surface is unknown, it can be predicted using various approaches. The energetic method computationally analyses the...
Binding sites are often located in large pockets, and if their location on a protein’s surface is unknown, it can be predicted using various approaches. The energetic method computationally analyses the...

