EGG

EGG assesses the accuracy of predicted multimeric protein models using deep energy-based models and graph neural networks to infer overall fold and interface accuracy scores.

Key Features:

• Deep energy-based models: EGG employs energy-based architectures to evaluate multimeric protein model quality.
• Graph neural networks: EGG implements message-passing and transformer layers to process protein structure representations.
• Overall fold accuracy estimation: EGG predicts the global structural quality of multimeric protein models.
• Interface accuracy estimation: EGG evaluates the quality of protein-protein interfaces in multimeric complexes.
• Top model ranking: EGG identifies the highest-quality models and top three highest-quality models from prediction sets.

Scientific Applications:

• Multimeric protein model quality assessment: EGG enables reliable estimation of accuracy for predicted protein complex structures to determine their utility.
• Quaternary structure analysis: EGG supports understanding of protein complex conformations relevant to cell biology, systems biology, disease formation, and disease treatment.
• CASP evaluation benchmarking: EGG provides validated performance for comparing and selecting among predicted multimeric models in community assessments.

Methodology:

EGG implements message-passing and transformer layers within deep energy-based and graph-based architectures to infer overall fold and interface accuracy scores of predicted multimeric protein models. Evaluation was performed against CASP15 targets.