Inferring sparse structure in genotype-phenotype maps.

Area of Science:

• Genetics
• Systems Biology
• Bioinformatics

Background:

• Phenotypic correlations in related individuals suggest shared genetic architecture.
• Pleiotropy, where single genetic loci affect multiple phenotypes, is a key driver of these correlations.
• A hypothesis posits that pleiotropic effects stem from a limited number of core cellular processes.

Purpose of the Study:

• To propose a novel computational method for inferring latent structure in genotype-phenotype data.
• To identify core cellular processes that underlie observed genotype-phenotype relationships.
• To leverage sparsity as a guiding principle for uncovering biological structure.

Main Methods:

• Developed Sparse Structure Discovery (SSD), a penalized matrix decomposition approach.
• SSD is designed to identify low-dimensional, locus-sparse, and/or phenotype-sparse latent structures.
• Validated SSD using synthetic data and applied it to empirical datasets from yeast and human cell lines.

Main Results:

• SSD accurately recovers core processes under conditions of locus or phenotype sparsity.
• Empirical application to yeast and human cell line data revealed biologically plausible core processes.
• Demonstrated evidence of sparse structure in existing genotype-phenotype datasets.

Conclusions:

• Sparsity is a valuable guiding prior for resolving latent structure in genotype-phenotype mapping.
• The proposed SSD method provides a robust framework for uncovering the genetic basis of complex traits.
• This approach advances our understanding of pleiotropy and its role in biological systems.