Breeding and Genetics Symposium: really big data: processing and analysis of very large data sets

Area of Science:

• Animal Breeding and Genetics
• Bioinformatics
• Computational Biology

Background:

• Modern animal breeding relies on large datasets from SNP arrays and sequencing.
• High-performance computing is essential for warehousing and analyzing this data.
• Genomic selection offers significant accuracy gains, particularly for low heritability traits.

Purpose of the Study:

• To explore computational challenges and solutions for large-scale genomic data analysis in animal breeding.
• To highlight the impact of advanced computing methods on genetic evaluation accuracy and speed.
• To discuss the integration of genomic data into breeding programs.

Main Methods:

• Development of high-performance computing methods for data warehousing and analysis.
• Application of efficient algorithms for matrix inversion and variance component estimation.
• Implementation of single-step approaches combining BLUP with genomic relationship matrices.
• Optimization of algorithms using custom libraries and parallel computing for constructing genomic matrices.

Main Results:

• Genomic selection has yielded greatest accuracy gains for low heritability traits.
• Efficient algorithms have reduced the time for constructing genomic matrices from hours to minutes.
• Single-step genomic evaluations show comparable computational demands to traditional BLUP.
• Large datasets necessitate robust data mining and visualization tools for insights.

Conclusions:

• Advancements in computing and algorithms are crucial for managing large genomic datasets in animal breeding.
• Optimized computational approaches accelerate genetic evaluations and genetic improvement rates.
• Effective data management and software engineering are key to realizing the full potential of genomic selection.