SampleExplorer

SampleExplorer embeds sample metadata using a transformer-based language model to enable researchers to search for relevant transcriptome datasets with text and gene set queries.

Key Features:

• Dual Query Search: Supports both text-based and gene set-based queries for dataset retrieval.
• Metadata Embedding: Embeds sample metadata using a transformer-based language model to capture semantic relationships.
• Similarity Retrieval: Retrieves biologically similar datasets from large-scale transcriptomic repositories based on embedded metadata representations.
• ARCHS4 Database Integration: Benchmarked and validated using the ARCHS4 database for retrieving biologically relevant samples.

Scientific Applications:

• Dataset Discovery: Enables efficient discovery of relevant gene expression datasets in large public repositories.
• Replication and Verification Studies: Helps researchers leverage existing transcriptomic data for potential replication or verification of published findings.
• Cross-Context Sample Identification: Improves sample and dataset identification across diverse experimental contexts in biomedical research.

Methodology:

SampleExplorer uses a transformer-based language model to embed sample metadata, with detailed algorithmic description of the retrieval process and data preparation steps provided in supplementary materials.