Experimentally Tuned Protein-RNA Rosetta Score Function using Bayesian Optimization

Area of Science:

• Biochemistry
• Computational Biology
• Structural Biology

Background:

• Protein-RNA complexes are crucial for essential cellular functions, including transcription and translation.
• Quantifying the energetic favorability of protein-RNA interactions is challenging due to a lack of reliable and accessible methods.

Purpose of the Study:

• To develop an experimentally tuned protein-RNA scoring function for the ROSETTA molecular modeling suite.
• To efficiently optimize scoring functions for protein-class interactions using Bayesian Optimization.

Main Methods:

• Implementation of a novel protein-RNA scoring function within ROSETTA.
• Application of Bayesian Optimization to fine-tune scoring function parameters for improved energetic agreement with experimental data.

Main Results:

• The developed scoring function demonstrates improved energetic prediction of protein-RNA interactions.
• Significant interactions were identified for specific RNA subclasses, validating the score function's physical basis.
• Bayesian Optimization significantly reduced the computational cost of fine-tuning ROSETTA scoring functions.

Conclusions:

• The study presents a validated, experimentally tuned scoring function for predicting protein-RNA interactions in ROSETTA.
• The proposed Bayesian Optimization framework offers an efficient method for optimizing ROSETTA scoring functions for diverse protein-class interactions.
• This work provides a foundation for more accurate computational studies of biomolecular interactions.