Accurate and Efficient KIR Gene and Haplotype Inference From Genome Sequencing Reads With Novel K-mer Signatures

Abstract

The killer-cell immunoglobulin-like receptor (KIR) proteins evolve to fight viruses and mediate the body’s reaction to pregnancy. These roles provide selection pressure for variation at both the structural/haplotype and base/allele levels. At the same time, the genes have evolved relatively recently by tandem duplication and therefore exhibit very high sequence similarity over thousands of bases. These variation-homology patterns make it impossible to interpret KIR haplotypes from abundant short-read genome sequencing data at population scale using existing methods. Here, we developed an efficient computational approach for in silico KIR probe interpretation (KPI) to accurately interpret individual’s KIR genes and haplotype-pairs from KIR sequencing reads. We designed synthetic 25-base sequence probes by analyzing previously reported haplotype sequences, and we developed a bioinformatics pipeline to interpret the probes in the context of 16 KIR genes and 16 haplotype structures. We demonstrated its accuracy on a synthetic data set as well as a real whole genome sequences from 748 individuals from The Genome of the Netherlands (GoNL). The GoNL predictions were compared with predictions from SNP-based predictions. Our results show 100% accuracy rate for the synthetic tests and a 99.6% family-consistency rate in the GoNL tests. Agreement with the SNP-based calls on KIR genes ranges from 72%–100% with a mean of 92%; most differences occur in genes KIR2DS2, KIR2DL2, KIR2DS3, and KIR2DL5 where KPI predicts presence and the SNP-based interpretation predicts absence. Overall, the evidence suggests that KPI’s accuracy is 97% or greater for both KIR gene and haplotype-pair predictions, and the presence/absence genotyping leads to ambiguous haplotype-pair predictions with 16 reference KIR haplotype structures. KPI is free, open, and easily executable as a Nextflow workflow supported by a Docker environment at https://github.com/droeatumn/kpi.

Increased risk of severe clinical course of COVID-19 in carriers of HLA-C*04:01

“…Exome data was aligned using BWA-MEM to the reference GRCh37 (hs357d.fa) [ 53 ]. Samples were identified as KIR2DS4f-positive [ 54 , 55 ] by screening the GATK HaplotypeCaller v3.7 [ 23 ] variant calls for presence of an insertion of CCCGGAGCTCCTATGACATGTA in exon 4 of KIR2DS4. For a more detailed description of the analysis please see methods in the data supplement.…”

HLA-C*04:01 is a Genetic Risk Allele for Severe Course of COVID-19

“…Unlike HLA, no ready-made bioinformatics tool is available for accurately haplotyping the genes in the KIR family, only KPI 49 , which requires whole genome sequencing data. We thus limited the analysis of the functional variant of KIR2DS4 50 being present (KIR2DS4f+) in the data as follows: The non-functional variant is the major allele of KIR2DS4 that is present in the human reference GRCh37.…”

KIR gene content imputation from single-nucleotide polymorphisms in the Finnish population

“…To compare the level of overall accuracy of our SNP-based method with an established sequencing-based approach, we extracted the results of the evaluation by Chen and coworkers (Chen et al, 2020) for the KIR imputation method kpi (Roe and Kuang, 2020). Figure 2b shows the accuracy of kpi compared with our test set results.…”

“…Another noteworthy limitation is that the method is not capable of identifying alleles. To date, only targeted sequencing based approaches can resolve KIR alleles (Maniangou et al, 2017; Norman et al, 2016; Roe and Kuang, 2020; Wagner et al, 2018). A possible future direction therefore is to extend KIR imputation from SNPs to cover allelic diversity.…”

“…Furthermore, in organ or stem cell transplantation setting the KIR locus offers additional genetic information for donor selection and prediction of clinical outcome (Cooley et al, 2010; Impola et al, 2014; Littera et al, 2017), and for many of these clinical genome datasets SNP microarray provides the most costeffective genotyping platform as well. To date, several KIR copy number or gene content analysis methods have been implemented for sequencing data (Chen et al, 2020; Maniangou et al, 2017; Norman et al, 2016; Roe and Kuang, 2020; Wagner et al, 2018), but to our knowledge only one SNP-based approach exists so far (Vukcevic et al, 2015). These approaches reach a high accuracy which makes KIR inference reliable enough for research and even practical clinical applications.…”