Alembic: a framework for converting disparate biological data into structured resources

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Public sequencing data repositories contain vast biological information.
• Data heterogeneity and non-standardized metadata impede effective data retrieval and analysis.
• Existing tools lack integrated workflows for accessing and processing diverse sequencing datasets.

Purpose of the Study:

• To develop a comprehensive solution for overcoming data heterogeneity in public sequencing repositories.
• To enable systematic search, integration, and comparative analysis of sequencing data.
• To streamline the workflow from data discovery to local repository construction.

Main Methods:

• Utilized the Entrez database system and its API for programmatic access to sequencing data and metadata.
• Applied Natural Language Processing (NLP) techniques, specifically transformer-based AI algorithms (PubMedBERT via AIONER platform), to analyze biomedical text.
• Developed the Alembic software package, a client-server system for automated data structuring and analysis.

Main Results:

• Transformed unstructured textual metadata into structured, computable information.
• Enabled efficient keyword-based searching and identification of relevant datasets, including gene-specific searches.
• Provided a curated list of datasets with structured metadata and direct links for downloading primary sequencing files.

Conclusions:

• Alembic offers a universal solution to the fragmented approach of existing tools for public sequencing data.
• The software facilitates efficient identification of high-value targets and construction of tailored local repositories.
• Alembic enhances the accessibility and utility of accumulated public sequencing data for biological research.