Graph-KIR: graph-based KIR copy number estimation and allele calling using short-read sequencing data

Area of Science:

• Genomics
• Immunogenetics
• Bioinformatics

Background:

• Killer-cell Immunoglobulin-like Receptor (KIR) genes are highly polymorphic and crucial in autoimmune diseases and transplantation.
• KIR gene sequences exhibit high similarity, complicating copy number estimation and high-resolution allele typing.
• Existing methods struggle with accurate KIR gene copy number and allele determination from whole genome sequencing (WGS) data.

Purpose of the Study:

• Introduce Graph-KIR, a novel computational tool for KIR gene analysis.
• Enable accurate estimation of KIR gene copy numbers from WGS data.
• Enable precise prediction of full-resolution (7-digit) KIR alleles from WGS data.

Main Methods:

• Developed Graph-KIR, a tool leveraging graph-based approaches for KIR gene analysis.
• Utilized simulated datasets for performance evaluation of copy number estimation and allele typing.
• Validated Graph-KIR performance on real-world Human Pangenome Reference Consortium (HPRC) samples.

Main Results:

• Graph-KIR achieved 99.2% accuracy in copy number estimation on simulated data.
• Demonstrated high F1-scores for allele typing: 91.79% (7-digit), 97.37% (5-digit), and 97.11% (3-digit).
• Outperformed existing tools (Geny, PING, T1K) in both simulated and HPRC sample analyses, especially at 7-digit resolution.

Conclusions:

• Graph-KIR provides a significant advancement in accurately analyzing KIR gene copy numbers and alleles from WGS data.
• The tool offers superior performance compared to current methods, enhancing KIR typing capabilities.
• Graph-KIR is a valuable resource for researchers studying KIR in various biological and clinical contexts.