Minimizing Reference Bias: The Impute-First Approach for Personalized Genome Analysis

We introduce the Impute-first alignment framework that reduces reference bias in genomics by integrating genotype imputation with pangenome alignment. Beginning with genotyping and genotype imputation using a portion of the input data, a personalized diploid reference genome is created. This personalized reference genome is then used for read alignment, enhancing downstream results such as variant calls. Our approach outperforms the traditional graph-based pangenome references in variant-calling measures, exhibiting improved sensitivity, precision, and overall accuracy, as demonstrated using the HG001 sample. The Impute-first framework successfully merges the benefits of traditional reference-based methods and pangenome strategies, reducing reference bias while maintaining a low computational cost.