Selecting among three basic fitness landscape models: Additive, multiplicative and stickbreaking.

Area of Science:

• Evolutionary Biology
• Genetics
• Computational Biology

Background:

• Fitness landscapes connect genotypes to organismal fitness, with topography influenced by epistasis.
• Understanding these landscapes is vital for evolutionary processes like adaptation, speciation, and recombination.
• Empirical testing of landscape models is increasingly feasible due to growing datasets.

Purpose of the Study:

• To develop statistical and computational methods for fitting fitness data to basic models.
• To establish a framework for comparing complex models against simpler, meaningful ones.
• To provide tools for analyzing empirical fitness data and understanding evolutionary dynamics.

Main Methods:

• Developed methods for fitting mutation combinatorial network data to additive, multiplicative, and stickbreaking models.
• Employed a Bayesian framework for robust model selection.
• Utilized simulations to assess statistical performance, including bias, error, and model discrimination power.

Main Results:

• Demonstrated the efficacy of the developed methods through simulations.
• Quantified the statistical performance of the methods in terms of bias, error, and discriminatory power.
• Successfully applied the approach to analyze several previously published datasets, showcasing flexibility.

Conclusions:

• The developed methods provide a robust framework for analyzing fitness landscapes and comparing evolutionary models.
• The R-package 'Stickbreaker' offers a practical tool for researchers to implement these methods.
• This work facilitates a deeper understanding of evolutionary processes by enabling better empirical testing of landscape models.