mouse-embeddings

mouse-embeddings maps high-dimensional phenotypic data from mouse models into low-dimensional embeddings to facilitate analysis of phenotypic diversity and relationships across genotypes.

Key Features:

• Dataset scope: Embeddings are derived from over 53,000 mouse models with mutations in more than 15,000 genomic markers and include over 254,000 phenotype annotations categorized by more than 9,000 ontology terms.
• Dimensional reduction and embedding techniques: The tool employs dimensional reduction methods to generate low-dimensional embeddings and visual maps that represent complex phenotypic relationships.
• Data integration and preprocessing: The workflow emphasizes preprocessing, filtering, and encoding strategies to address sparse phenotypic data and influence the resulting embeddings.
• Exploratory analysis capabilities: Embeddings enable exploratory searches to identify mouse models and phenotypic patterns relevant for studying gene function and human disease models.

Scientific Applications:

• Gene function studies: Visual embeddings support analysis of how specific genes influence phenotype across a broad collection of mouse models.
• Disease modeling: Embedding-based similarity and clustering aid identification of mouse models that recapitulate aspects of human diseases for translational research.
• Phenotypic data exploration: Low-dimensional representations facilitate discovery of novel phenotypic patterns and relationships among genotype-phenotype annotations.

Methodology:

Computational steps explicitly include dimensional reduction to produce embeddings and visual maps, together with data preprocessing, filtering, and encoding to handle sparse phenotypic annotations, implemented via analysis scripts.