Exploring imputation performance of phenotype-associated SNPs for forensic prediction models

Area of Science:

• Genetics
• Forensic Science

Background:

• Imputing single nucleotide polymorphisms (SNPs) is common in medical genetics but underutilized in forensics.
• Forensic DNA phenotyping relies on SNP-based phenotype prediction, which can be improved by imputing missing SNPs.

Purpose of the Study:

• Compare imputation accuracy for phenotype prediction SNPs, phenotype-associated SNPs, and random SNPs.
• Evaluate the performance of the HIrisPlex-S phenotype prediction model using imputed datasets.

Main Methods:

• Compared imputation accuracy across different SNP sets (phenotype prediction, phenotype-associated, random).
• Assessed the impact of SNP selection, dataset size, and minor allele frequencies (MAFs) on imputation.
• Evaluated phenotype prediction model performance with imputed genotype data.

Main Results:

• SNP selection and MAFs significantly influence imputation call and error rates.
• Phenotypic SNPs exhibit higher imputation error rates than random SNPs due to MAF differences.
• Imputation errors have less impact on trait prediction than missing genotypes, supporting lenient imputation thresholds.

Conclusions:

• Imputation performance studies for specific SNPs are vital for forensic applications.
• SNP imputation is highly applicable and beneficial prior to forensic phenotype prediction.
• Accurate phenotype prediction is achievable with dense SNP panels, frequent phenotypes, and appropriate imputation strategies.