GenomicSuperSignature

GenomicSuperSignature interprets new transcriptomic datasets through near-instantaneous comparison to public RNA-seq archives without high-performance computing requirements.

Key Features:

• Pre-computed Index: A pre-computed index enables rapid identification of public resources associated with a query dataset, including gene sets, MeSH terms, and publications.
• RAV Annotation: Replicable Axes of Variation (RAVs) are annotated with metadata from originating studies and by gene set enrichment analysis.
• Visualization: Functions for intuitive visualization of results are implemented within the package.
• Batch Effect Robustness: The method demonstrates robustness to batch effects and heterogeneous training data.
• Transfer Learning: The method supports transfer learning capacity for application to diverse datasets.

Scientific Applications:

• RNA-seq Interpretation: Analyzing new gene expression data in the context of existing transcriptomic databases using minimal computing resources.
• Database Search: Efficient and coherent search across large public transcriptomic archives for related studies and biological patterns.
• Cancer Genomics: Demonstrated using TCGA datasets for interpreting tumor transcriptomic profiles.
• Rare Disease Analysis: Applied to rare diseases datasets to leverage public data for interpretation of limited or unusual cohorts.

Methodology:

Principal Component Analysis is applied on 536 studies comprising 44,890 human RNA sequencing profiles, and sufficiently similar loading vectors are aggregated to form Replicable Axes of Variation (RAVs).