Inference and visualization of complex genotype-phenotype maps with gpmap-tools.

Area of Science:

• Genetics and Genomics
• Computational Biology
• Molecular Biology

Background:

• Multiplex assays of variant effect (MAVEs) generate vast amounts of data on sequence variants.
• Understanding complex genotype-phenotype maps is hindered by a lack of analytical tools for high-dimensional MAVE data.
• Higher-order genetic interactions are prevalent but difficult to study with current methods.

Purpose of the Study:

• To develop and present gpmap-tools, a Python library for analyzing MAVE data and genotype-phenotype maps.
• To enable inference, imputation, and error estimation from complex, high-dimensional genetic datasets.
• To visualize and summarize patterns of epistasis and non-linear relationships in genotype-phenotype maps.

Main Methods:

• Development of the gpmap-tools Python library integrating Gaussian process models.
• Application of non-linear dimensionality reduction for visualizing large genotype-phenotype maps.
• Analysis of the Shine-Dalgarno sequence genotype-phenotype map using MAVE data and E. coli genomic sequences.

Main Results:

• gpmap-tools successfully infers and visualizes genotype-phenotype landscapes with millions of genotypes.
• Analysis of the Shine-Dalgarno sequence revealed a highly epistatic map.
• Inferred landscapes from genomic and experimental data showed strong consistency.

Conclusions:

• gpmap-tools provides essential tools for exploring complex genotype-phenotype relationships from MAVE data.
• The study uncovers a simple molecular mechanism governing the epistatic nature of the Shine-Dalgarno sequence.
• This work facilitates a deeper understanding of genetic interactions and sequence-function relationships.