An in-silico method for identifying aggregation rate enhancer and mitigator mutations in proteins

Area of Science:

• Biochemistry and Molecular Biology
• Computational Biology
• Biophysics

Background:

• Protein misfolding and aggregation are implicated in various diseases.
• Cellular quality control mechanisms can be overwhelmed by mutations, stress, or aging.
• Single amino acid substitutions can significantly alter protein aggregation rates.

Purpose of the Study:

• To develop a predictive model for mutation-induced protein aggregation.
• To identify key features distinguishing aggregation-enhancing from aggregation-mitigating mutations.
• To create a publicly available algorithm for predicting mutation effects on aggregation.

Main Methods:

• Collected and classified a dataset of 220 mutations across 25 proteins.
• Employed machine learning, specifically Support Vector Machines (SVM), for classification.
• Analyzed mutation effects in relation to local secondary structures (alpha-helices, beta-strands, coils).

Main Results:

• An initial SVM model achieved 69% accuracy in distinguishing mutation types.
• Incorporating local secondary structure information improved prediction accuracy by 13-15%.
• Identified distinct important features for mutations within different secondary structures (e.g., stability/flexibility in helices, propensity/charge in strands).

Conclusions:

• Sequence-based features, influenced by local secondary structure, can predict mutation effects on protein aggregation.
• A novel algorithm is available for predicting whether mutations enhance or mitigate aggregation.
• Applications include disease research, biopharmaceutical optimization, and bio-nanomaterial design.