Personal Precise Force Field for Intrinsically Disordered and Ordered Proteins Based on Deep Learning

Area of Science:

• Biochemistry
• Computational Biology
• Structural Biology

Background:

• Intrinsically disordered proteins (IDPs) lack fixed 3D structures and are implicated in various diseases.
• Experimental methods struggle to capture the dynamic conformations of IDPs.
• Molecular dynamics (MD) simulations offer a way to study IDP conformations, but accuracy relies on force fields.

Purpose of the Study:

• To develop improved force field parameters for accurately simulating intrinsically disordered proteins (IDPs).
• To enhance the accuracy of molecular dynamics (MD) simulations for IDPs by addressing limitations in current force fields.

Main Methods:

• Utilized deep learning to predict residue dihedral angles with higher accuracy.
• Employed reweighting techniques to optimize personalized force field parameters for individual residues.
• Evaluated the new force fields using intrinsically disordered proteins, structured proteins, and fast-folding proteins.

Main Results:

• The deep learning method demonstrated superior accuracy in predicting dihedral angles compared to previous approaches.
• Two personalized force field parameters, PPFF1 and PPFF1_af2, showed improved ability to reproduce experimental data.
• The developed force fields outperformed the standard ff03CMAP force field in simulations.

Conclusions:

• The developed strategy enables the creation of precise, personalized force fields for IDPs.
• This advancement is crucial for enhancing the accuracy of MD simulations and understanding IDP behavior in health and disease.
• The findings pave the way for more reliable computational studies of proteins involved in Parkinson's, Alzheimer's, and cancer.