Accelerating imputation of missing genotypes using parallel computing

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Advancements in DNA technology yield large-scale genomic datasets, posing computational challenges for analysis.
• Current computational tools struggle with the size of these datasets, leading to lengthy analysis times.

Purpose of the Study:

• To investigate the impact of parallel computing on the random forest (RF) algorithm's performance for imputing missing genotypes.
• To assess the effect of parallel computing on both the accuracy and computational time of genotype imputation.

Main Methods:

• Simulated genotypic matrices with varying numbers of single-nucleotide polymorphisms (SNPs) and individuals.
• Masked 50% of genotypic data and imputed using the RF algorithm under serial and parallel computing conditions.
• Evaluated imputation accuracy by the percentage of correctly imputed genotypes and measured computational time.

Main Results:

• Parallel computing did not alter the accuracy of genotype imputation compared to serial computing.
• Parallel computing significantly reduced analysis time, decreasing running time by up to 63%.
• CPU utilization increased from 10% in serial computing to 55% in parallel computing.

Conclusions:

• Parallel computing offers a substantial speed-up for analyzing large genomic datasets with the RF algorithm for genotype imputation.
• Researchers should leverage parallel computing to efficiently analyze large genomic datasets without compromising imputation accuracy.