KPI

KPI infers killer-cell immunoglobulin-like receptor (KIR) gene presence/absence and haplotype-pair assignments from short-read whole-genome sequencing data using k-mer–based probe signatures.

Key Features:

• Novel k-mer signatures: Leverages novel k-mer signatures to distinguish highly similar KIR gene regions arising from tandem duplications.
• Synthetic 25-base sequence probes: Uses synthetic 25-base probes designed from previously reported haplotype sequences to target 16 distinct KIR genes and corresponding haplotype structures.
• Bioinformatics pipeline: Interprets probe hits in sequencing reads to determine individual KIR gene configurations and infer haplotype-pair associations.
• Validation on synthetic and GoNL WGS: Achieved 100% accuracy on synthetic tests and 99.6% family-consistency on 748 whole genomes from The Genome of the Netherlands (GoNL).
• Comparison to SNP-based methods: Shows 72–100% agreement (mean 92%) with SNP-based predictions, with primary discrepancies in KIR2DS2, KIR2DL2, KIR2DS3, and KIR2DL5.
• Accuracy for gene and haplotype-pair calls: Reported overall accuracy of 97% or higher for both gene and haplotype-pair predictions using a reference set of 16 KIR haplotypes.

Scientific Applications:

• Immunogenetics: Enables study of KIR genetic variation that influences antiviral immune responses and reproductive immunology.
• Population-scale KIR genotyping: Facilitates large-scale inference of KIR gene and haplotype frequencies from short-read whole-genome sequencing data.
• Association and clinical studies: Supports investigations of KIR-related disease susceptibility and treatment outcome associations through accurate gene and haplotype calls.

Methodology:

Designs synthetic 25-base probes from reported haplotype sequences and uses novel k-mer signatures; searches probe sequences in short-read sequencing data and applies a bioinformatics pipeline to call gene presence/absence and infer haplotype pairs; validated against synthetic datasets and 748 GoNL whole genomes and compared to SNP-based predictions.