AltGosling: automatic generation of text descriptions for accessible genomics data visualization.

Area of Science:

• Genomics
• Bioinformatics
• Data Visualization

Background:

• Biomedical visualizations are crucial for knowledge discovery and pattern detection in large datasets.
• Interactive visualizations are vital for data scientists but inaccessible to visually impaired individuals without proper descriptions.
• Current alt text generation models are inadequate for complex genomics data visualizations.

Purpose of the Study:

• To develop an automated tool for generating descriptions of interactive genomics data visualizations.
• To enhance the accessibility of genomic data for people with blindness and low vision.
• To provide an alternative to manual description writing, which is often challenging and time-consuming.

Main Methods:

• Developed AltGosling, a logic-based algorithm for generating descriptions of genome-mapped data visualizations.
• Co-designed AltGosling with a blind screen reader user to ensure usability and effectiveness.
• Utilized the grammar-based genomics toolkit Gosling for visualization creation.

Main Results:

• AltGosling generates diverse descriptions, including a navigable tree-structured panel.
• The tool significantly outperforms state-of-the-art large language models and image-based neural networks in generating alt text for genomics visualizations.
• AltGosling represents a novel solution for increasing accessibility in genomic research.

Conclusions:

• AltGosling effectively addresses the challenge of describing complex genomics visualizations.
• The tool enhances data accessibility for visually impaired researchers and scientists.
• AltGosling sets a precedent for developing accessible visualization tools in bioinformatics and genomics.