MELD × MD Folds Nonthreadables, Giving Native Structures and Populations

Area of Science:

• Computational biology
• Biophysics
• Structural biology

Background:

• Protein structure prediction from amino acid sequences is a key challenge.
• Threading is a common bioinformatics approach but fails for ~15% of proteins (nonthreadables).
• Physical molecular dynamics (MD) modeling offers a template-free alternative for nonthreadable proteins.

Purpose of the Study:

• To evaluate the efficacy of MELD-accelerated MD (MELD × MD) in folding nonthreadable proteins.
• To assess the performance of MELD × MD for proteins with fewer than 125 residues.
• To investigate MELD × MD's ability to predict its own success.

Main Methods:

• Utilizing MELD (Molecular Enhanced Sampling with Local DEcomposition) to accelerate MD simulations.
• Applying MELD × MD to 41 nonthreadable proteins with <125 residues.
• Analyzing simulation results, including accuracy (error <4 Å) and convergence (large Boltzmann populations).

Main Results:

• MELD × MD successfully folded 20 out of 41 nonthreadable proteins to within a 4 Å error.
• The method succeeded in 10 cases even when the force field had inaccuracies.
• In 11 cases, simulation convergence (large Boltzmann populations) predicted successful folding.

Conclusions:

• MELD-accelerated MD is a powerful tool for predicting the native structures of nonthreadable proteins.
• This method expands the range of physical protein modeling, addressing limitations of threading.
• MELD × MD demonstrates potential for reliable protein structure prediction, even in challenging cases.