GWAS-Flow

GWAS-Flow is a software tool developed to address the computational demands of modern genome-wide association studies (GWAS) in the context of identifying associations between complex traits and genomic variations. As genotyping and phenotyping large populations have become more feasible, handling vast amounts of data in GWAS has become a challenge. This is particularly true when implementing permutation-based significance thresholds, which offer advantages in adjusting for multiple hypotheses and non-Gaussian phenotypic distributions. GWAS-Flow leverages the machine learning framework TensorFlow to create a linear mixed model capable of efficiently analyzing large datasets by utilizing available CPU or GPU resources, thereby significantly reducing analysis time.

The tool offers notable advantages over custom GWAS scripts in terms of both speed and accuracy. GWAS-Flow not only calculates p-values but also generates essential summary statistics such as effect size and standard error for each marker. The tool provides a CPU-based option suitable for smaller datasets, while a GPU-based version is optimized for analyzing larger datasets. By combining TensorFlow's capabilities with its efficient approach, GWAS-Flow enhances the computational efficiency of GWAS, making it more manageable for researchers dealing with extensive datasets.