Gemma

Gemma curates and standardizes gene expression datasets for reuse and meta-analysis of microarray and RNA-seq experiments.

Key Features:

• Data curation and content: A curated database containing 10,811 manually curated datasets (June 2020) encompassing over 395,000 samples from human, mouse, and rat and hundreds of transcriptomic platforms for microarray and RNA sequencing technologies.
• Semantic annotation: Extraction of concepts from natural language using the Unified Medical Language System (UMLS) and linking annotations to classes in open biomedical ontologies.
• Annotation quality: Predicted semantic annotations achieve 89% precision after manual evaluation and correction.
• Data processing and quality control: Standardized processing and quality-control procedures addressing inconsistent probe-gene mappings and unstructured metadata to support reliable analyses.
• Differential expression analysis: Provision of standardized differential expression results across diverse datasets to enable comparative and integrative analyses.
• Coverage and ontology representation: Representation of dataset topics using 10,215 distinct terms from 12 ontologies, yielding 54,316 topic annotations, with nervous system-related experiments comprising 34% of holdings.

Scientific Applications:

• Meta-analysis: Enabling cross-study aggregation to identify consistent expression patterns and replicate findings across multiple experiments.
• Differential expression analysis: Supporting standardized identification of differentially expressed genes across conditions and platforms.
• Neuroscience research: Providing extensive brain- and nervous system-related datasets for comparative and integrative studies.

Methodology:

Re-analysis of raw data from public sources such as NCBI GEO, annotation of experimental conditions using UMLS-derived concepts and ontology links, implementation of standardized quality control, and integration of semantic annotation pipelines.