ChatMDV: reducing technical barriers in bioinformatics analysis using large language models

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Advanced omics technologies generate complex biological data.
• Existing visualization tools like Multi-Dimensional Viewer (MDV) require significant computational expertise.
• Limited accessibility hinders widespread use of powerful data exploration platforms.

Purpose of the Study:

• To develop a natural language interface for MDV to simplify biological data visualization.
• To enhance accessibility and reproducibility in analyzing complex omics datasets.
• To enable users to generate interactive visualizations and analyses via conversational commands.

Main Methods:

• Integration of a natural language interface with the Multi-Dimensional Viewer (MDV).
• Utilized a retrieval-augmented generation (RAG) pipeline and large language models (LLMs).
• Translated natural language queries into executable Python code for reproducible analyses.

Main Results:

• ChatMDV successfully generated high-quality, interactive visualizations from natural language queries.
• Demonstrated capabilities across diverse datasets including single-cell RNA-sequencing (scRNA-seq) data.
• Achieved high semantic success rates (79-97%) in data visualization tasks.

Conclusions:

• ChatMDV bridges natural language processing and bioinformatics visualization, reducing technical barriers.
• Enhances reproducibility and inclusivity in scientific inquiry.
• Its modular and FAIR-compliant design supports scalable biological data analysis.