Delivering artificial intelligence-ready genomics with the Maize Genetics and Genomics Database

Area of Science:

• Computational biology
• Agricultural biotechnology
• Genomics

Background:

• Artificial Intelligence (AI) integration is transforming biological research, especially in agriculture.
• Maize (Zea mays L.) possesses extensive genomic and genetic resources, making it ideal for AI-driven crop improvement.
• Large, complex datasets in maize research present opportunities for AI-based discovery.

Purpose of the Study:

• To detail MaizeGDB's strategic initiatives in creating an AI-ready infrastructure for maize data.
• To enable the maize research and breeding community to leverage AI for accelerated discovery.
• To standardize datasets and generate precomputed embeddings using DNA and protein language models.

Main Methods:

• Standardizing maize datasets and pre-computing complex features.
• Developing novel interactive tools and reproducible workflows.
• Generating precomputed embeddings from DNA and protein language models for AI applications.

Main Results:

• MaizeGDB has established an AI-ready data foundation in standardized formats.
• New functionalities include zero-shot variant effect scoring and visualization of nucleotide conservation.
• Custom dataset assembly resources and reproducible workflows are available via GitHub.

Conclusions:

• MaizeGDB's AI-ready infrastructure facilitates AI integration in maize research.
• This initiative accelerates the discovery of gene function and variant interpretation in maize.
• The resources provided will aid in the development of improved maize varieties through AI.