BioVDB: biological vector database for high-throughput gene expression meta-analysis.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• High-throughput sequencing generates vast amounts of gene expression data, often publicly available in repositories like NCBI's Gene Expression Omnibus (GEO).
• Analyzing and querying this exponentially growing dataset for patterns like similarity and distance presents significant challenges.
• Vectorization of gene expression data is a common prerequisite for Artificial Intelligence (AI) and Machine Learning (ML) applications.

Purpose of the Study:

• To introduce BioVDB, a novel vector database for the efficient storage and analysis of gene expression data.
• To enhance the integration of biological studies with AI/ML tools.
• To facilitate pattern discovery and similarity analysis within large-scale gene expression datasets.

Main Methods:

• Development and implementation of BioVDB, a vector database architecture.
• Utilization of Automatic Label Extraction (ALE) to extract sample metadata labels (age, sex, tissue/cell-line).
• Ingestion of 438,562 samples from eight microarray GEO platforms into BioVDB.

Main Results:

• BioVDB enables efficient querying of gene expression data through similarity search.
• Demonstrated utility in identifying and inferring missing sample labels.
• Facilitates rapid similarity analysis across a large number of samples.

Conclusions:

• BioVDB provides a scalable solution for managing and analyzing large gene expression datasets.
• The database enhances the potential for leveraging AI/ML techniques in biological research.
• BioVDB supports efficient data exploration, label inference, and similarity assessments.