A comparative analysis of current phasing and imputation software

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Whole-genome data accessibility has surged due to reduced sequencing costs and imputation methods.
• Microarray chips combined with imputation enable near whole-genome data generation affordably.
• Hidden Markov Models (HMM) are the predominant approach for imputation.

Purpose of the Study:

• To compare the performance of leading HMM-based phasing and imputation tools.
• To evaluate accuracy, speed, and memory usage across different datasets and chip densities.
• To guide users in selecting optimal software and reference panels for imputation.

Main Methods:

• Benchmarking of Beagle5.4, Eagle2.4.1, Shapeit4, Impute5, and Minimac4.
• Testing on four datasets with varying microarray chip densities.
• Assessment of imputation accuracy using concordance rates, IQS, and R2 metrics.
• Evaluation of computational resources including run time and memory usage.

Main Results:

• Beagle5.4 achieved the highest average concordance rate, followed by Impute5 and Minimac4.
• Impute5 and Minimac4 excelled with low-frequency markers, while Beagle5.4 was more accurate for common markers (MAF>5%).
• Beagle5.4 demonstrated lower run times; Minimac4 utilized the least memory among imputation tools.

Conclusions:

• The choice of imputation tool and metric impacts interpretation of results.
• Specific software combinations are best suited for different genomic analysis needs.
• An automated pipeline was developed to assist users in creating customized imputation strategies.