A physics-based energy function allows the computational redesign of a PDZ domain.

Area of Science:

• Biochemistry
• Structural Biology
• Computational Biology

Background:

• Computational protein design (CPD) aims to solve the inverse folding problem by predicting sequences that fold into desired structures.
• Previous CPD efforts utilized knowledge-based energy functions for both folded and unfolded protein states.
• The PDZ domain is a common structural motif in signaling proteins.

Purpose of the Study:

• To investigate the feasibility of entirely redesigning a PDZ domain using a hybrid energy function.
• To evaluate the efficacy of combining physics-based and knowledge-based energy terms in CPD.
• To experimentally validate computationally designed protein sequences.

Main Methods:

• Employed a hybrid energy function: physics-based for the folded state and knowledge-based for the unfolded state.
• Generated thousands of candidate sequences using Monte Carlo simulations.
• Selected three top-scoring sequences for experimental characterization based on energy and empirical criteria.

Main Results:

• All three redesigned sequences were successfully overexpressed in the laboratory.
• Circular dichroism and 1D-NMR spectra indicated native-like folding for all tested constructs.
• Two constructs exhibited ligand-induced shifts in thermal denaturation, confirming correct folding and function.

Conclusions:

• Successful whole-protein redesign of a PDZ domain is achievable using a physics-based energy function for the folded state and a knowledge-based function for the unfolded state.
• The combination of physical principles and empirical filtering is a viable strategy for de novo protein design.
• This study validates the power of computational approaches in creating functional protein structures.