Higher-order epistasis and phenotypic prediction.

Area of Science:

• Genetics
• Computational Biology
• Molecular Biology

Background:

• High-throughput mutagenesis experiments reveal complex genetic interactions beyond pairwise effects.
• Understanding higher-order genetic interactions (epistasis) is crucial but challenging for modeling genotype-phenotype relationships.

Purpose of the Study:

• To develop a method for reconstructing sequence-to-function mappings that accommodates all orders of genetic interaction.
• To predict unobserved genotypes by accurately reflecting the type and extent of epistasis present in observed data.

Main Methods:

• Extracting epistasis information by analyzing changes in phenotypic correlations with mutational distance.
• Estimating variance components attributed to different orders of genetic interaction (additive, pairwise, three-way, etc.).
• Employing an empirical Bayes prior and Gaussian process regression to reconstruct genotype-phenotype mappings.

Main Results:

• The method successfully reconstructs sequence-to-function mappings, accommodating all orders of genetic interaction.
• Demonstrated application to the antibody-binding domain GB1.
• Validated predictions on human pre-mRNA splice site splicing efficiency data.

Conclusions:

• The developed method provides a powerful framework for modeling complex genetic interactions and genotype-phenotype relationships.
• Accurate reconstruction of sequence-to-function mappings is achievable even with partially observed data.
• The approach has broad applicability in understanding genetic architectures from mutagenesis experiments.