CATE

CATE accelerates large-scale molecular evolution analyses by implementing CUDA-accelerated parallel tests to detect selection and perform neutrality statistics on genomic datasets.

Key Features:

• Highly Parallelized Algorithms: Leverages CUDA on graphical processing units (GPUs) and central processing units (CPUs) to achieve large-scale parallelization of evolutionary tests.
• Novel File Organization and Search Algorithm: Implements a unique file organization system and an innovative search algorithm to optimize data handling and retrieval.
• Comprehensive Evolutionary Tests: Supports Tajima's D, Fu and Li's test statistics, Fay and Wu's test statistics, the McDonald–Kreitman Neutrality Index, Fixation Index, and Extended Haplotype Homozygosity.
• Performance and Scalability: CUDA-based implementation delivers high throughput and scalability for large genomic datasets, with benchmarked speedups over 180× compared to conventional methods.
• Adaptable GPU-Accelerated Framework: Framework can be applied to other large-scale parallel analyses in evolutionary and genomic research.

Scientific Applications:

• Selection and Neutrality Testing: Detects and quantifies selection pressures and departures from neutrality using established summary statistics and tests.
• Population Genomics: Enables analysis of population-scale datasets to study genetic variation and population-specific evolutionary patterns.
• Large-Scale Genomic Studies: Facilitates rapid processing of genome-wide datasets (for example, multi-population analyses of the 1000 Genomes Project) to increase statistical power in evolutionary inference.

Methodology:

The implementation uses CUDA to accelerate computations on GPUs while also leveraging CPUs, employs a novel file organization system and search algorithm for optimized data handling, and was benchmarked at over 180× speedup (processing all 54,849 human genes across the five super populations of the 1000 Genomes Project in under 30 minutes versus approximately 3.62 days with conventional methods).