The epigenetic dimension of protein structure

Area of Science:

• Structural biology
• Computational biology
• Biophysics

Background:

• Accurate protein structure prediction is a major challenge in biology.
• AlphaFold, a machine learning method, has claimed success in predicting structures for most human proteins.
• This claim warrants caution due to identified flaws in AlphaFold models.

Purpose of the Study:

• To critically evaluate the accuracy of AlphaFold predictions, particularly for membrane proteins and disordered regions.
• To propose a new paradigm, the "epigenetic dimension of protein structure," to account for factors influencing protein structure beyond the amino acid sequence.
• To reconcile the deterministic nature of genomic information with the inherent variability in biological systems.

Main Methods:

• Analysis of AlphaFold models to identify structural inaccuracies, focusing on disordered regions and membrane proteins.
• Conceptual framework development to define and incorporate the "epigenetic dimension of protein structure."
• Literature review and theoretical considerations on protein folding determinants.

Main Results:

• Identified significant flaws in AlphaFold models, including unrealistic representations of disordered regions and clashes in protein geometry.
• Observed particular inaccuracies in the prediction of membrane protein structures.
• Proposed the concept of the "epigenetic dimension of protein structure" encompassing non-sequence-encoded factors like lipid solvation and ligand binding.

Conclusions:

• AlphaFold's success should be tempered by its limitations in predicting complex protein structures.
• The "epigenetic dimension of protein structure" is a necessary concept to fully understand protein folding.
• This paradigm helps explain the balance between biological coding and inherent variability in living systems.