Importance of higher-order epistasis in protein sequence-function relationships

Area of Science:

• Biochemistry
• Computational Biology
• Genetics

Background:

• Protein sequence-function relationships are complex and difficult to predict.
• Pairwise epistasis is well-studied, but higher-order epistasis remains poorly understood and challenging to model.
• Accurate predictive models for novel genotypes are crucial for protein engineering.

Purpose of the Study:

• To develop and validate a novel transformer-based approach for modeling higher-order epistasis in proteins.
• To assess the significance of higher-order epistasis across diverse protein sequence-function datasets.
• To evaluate the impact of including higher-order epistasis on predictive model performance.

Main Methods:

• Introduced a transformer-based computational method to model adjustable orders of epistatic interactions.
• Incorporated global nonlinearity effects induced by experimental conditions.
• Applied the method to 10 large protein sequence-function datasets.

Main Results:

• The importance of higher-order epistasis varies significantly among proteins, explaining up to 60% of epistatic variance.
• Including higher-order epistasis enhances generalization of fitness data to distant sequence spaces.
• Higher-order epistasis is crucial for modeling complex, multi-peak fitness landscapes.

Conclusions:

• Higher-order epistasis plays a substantial role in protein sequence-function relationships.
• The developed transformer approach effectively models higher-order interactions.
• Incorporating higher-order epistasis is essential for advanced protein engineering and evolutionary data analysis.