# Variant Calling in the HLA Region: Overcoming High Polymorphism and Structural Variation


## Key Takeaways

- Standard variant calling pipelines fail in the HLA region due to extreme polymorphism, gene duplication, and structural variation, leading to multi-mapping reads that are either discarded or incorrectly assigned, resulting in missed true variants or false positives. Specialized tools are required to overcome these alignment ambiguities and reference biases.
- Accurate HLA variant calling necessitates specialized tools like HLA-HD or T1K, which directly genotype alleles rather than relying on reference-based alignment. HLA-HD demonstrates high concordance (e.g., 96.9% for allele genotyping), while T1K offers joint HLA and KIR genotyping, crucial for understanding immune system gene families.
- Sequencing depth is critical, with approximately 15-fold coverage for whole-genome sequencing providing a balance for accurate SNV calling and sufficient HLA allele genotyping. Higher depths are required for precise indel detection, and for whole-exome sequencing, adequate on-target depth at HLA loci is paramount.
- A practical workflow integrates standard genome-wide analysis with specialized HLA typing, emphasizing data quality assessment, careful alignment inspection, direct input of raw reads into HLA tools, and careful integration and validation of results, prioritizing specialized HLA calls in case of discrepancies.
- Reproducibility is achieved through containerized environments (e.g., Docker, Singularity), pinned tool versions, and structured pipeline parameters, ensuring consistent results across different computational platforms and over time, which is essential for the complex HLA region.
- Common failure patterns include low/zero coverage at specific HLA loci, an excess of false variants due to multi-mapping reads, and discordance between genome-wide and specialized HLA typing results, necessitating rigorous quality control at sample, read, alignment, and variant levels.

---

The human leukocyte antigen (HLA) region on chromosome 6 presents one of the most challenging targets for variant calling in human genomics. Standard variant calling pipelines that align short reads to the reference genome and identify differences frequently fail in this region because the extreme polymorphism, gene duplication, and structural variation create ambiguous alignments and false variant calls. This article provides a practical framework for researchers and laboratory professionals who need accurate variant detection in the HLA region, covering specialized typing tools, workflow design, quality control, and interpretation limits.

## The Biological Basis of HLA Region Complexity

The HLA region spans approximately 4 megabases on the short arm of chromosome 6 and contains the most polymorphic genes in the human genome. The classical HLA class I genes (HLA-A, HLA-B, HLA-C) and class II genes (HLA-DR, HLA-DQ, HLA-DP) encode proteins that present peptide antigens to T cells, and their diversity directly influences immune recognition. The clinical and research significance of this region extends beyond transplantation matching. HLA alleles are strongly associated with many immune-mediated and infection-related diseases, and the region's complex linkage disequilibrium patterns make traditional single nucleotide polymorphism association studies unreliable without specialized approaches.

The core problem for variant calling is that HLA genes exist in multiple highly similar copies. A sequencing read originating from one HLA gene can align almost equally well to several other HLA genes or to different alleles of the same gene. This sequence similarity creates what bioinformaticians call multi-mapping reads, which standard variant callers typically discard or assign arbitrarily. When reads are discarded, true variants are missed. When reads are assigned to the wrong location, false variants are introduced.

Structural variation compounds the problem. The HLA region contains copy number variations, insertions, deletions, and gene conversions that alter the local genomic architecture between individuals. A reference genome represents only one possible arrangement, and any individual whose HLA region differs structurally will show alignment artifacts that mimic true variants. The killer cell immunoglobulin-like receptor (KIR) genes, located nearby on chromosome 19, present an even more extreme version of this challenge because many KIR genes are similar to each other and may be entirely absent from some chromosomes.

## Why Standard Variant Calling Pipelines Fail in the HLA Region

Standard germline variant calling workflows follow a predictable sequence: align reads to the reference genome with a short-read aligner, mark duplicates, recalibrate base qualities, and call variants with a haplotype-based caller. These pipelines work well in unique regions of the genome but break down in the HLA region for several identifiable reasons.

The first failure point is read alignment. Short reads of 100 to 150 base pairs frequently cannot be uniquely placed when the reference contains multiple nearly identical HLA gene sequences. Aligners report a mapping quality near zero for these reads, and variant callers filter them out. The result is a coverage gap in the HLA genes, and variants in those gaps are never observed.

The second failure point is reference bias. The reference genome contains one specific HLA allele at each locus. Reads from an individual carrying a different allele will have more mismatches to the reference, and the alignment algorithm may place those reads suboptimally or reject them. This systematically reduces the ability to detect variants that differ from the reference allele.

The third failure point is the variant caller's assumption of a diploid genome with two copies of each locus. The HLA region's gene duplications and copy number variation violate this assumption. When a caller sees excess coverage at a locus, it may interpret the extra reads as evidence of a duplication or as sequencing artifacts, producing spurious variant calls or incorrect genotype assignments.

The fourth failure point is the interpretation of variants in the context of HLA alleles. A single nucleotide change in an HLA gene can alter the encoded amino acid and change the allele designation. Standard variant callers report individual variants but do not assemble them into the phased allele haplotypes that matter for HLA typing. Researchers need to know which variants occur together on the same chromosome, and standard callers do not provide this information reliably in the HLA region.

## Specialized HLA Typing Tools and Their Approaches

Because standard pipelines fail, the field has developed specialized tools that treat HLA genotyping as a distinct problem. These tools generally fall into two categories: those that assign alleles from sequencing data and those that perform graph-based or k-mer based genotyping. The choice of tool depends on the sequencing data type, the required resolution, and the downstream application.

The T1K method represents a modern approach that jointly considers alleles across all genotyped genes. This joint modeling allows the tool to reliably identify which genes are present and to distinguish homologous genes, including the challenging KIR2DL5A and KIR2DL5B pair. T1K works with RNA-seq, whole-genome sequencing, and whole-exome sequencing data, and it can call novel single nucleotide variants and process single-cell data. The ability to genotype KIR and HLA genes from the same data set is particularly valuable because both gene families are highly polymorphic and cannot be genotyped with standard variant calling pipelines.

The HLA-HD algorithm has demonstrated high accuracy in multiple benchmarking studies. In an evaluation using ultra-deep whole-genome sequencing data, HLA-HD achieved the highest concordance among the tools tested, with 96.9 percent concordance for HLA allele genotyping. In a large-scale application to whole-exome sequencing data from 454,824 UK Biobank participants, HLA-HD directly called HLA alleles and proved more accurate than imputing HLA alleles from genotyping arrays. The direct calling approach improved statistical power for rare alleles and for non-European ancestries, where imputation accuracy is lower.

Other tools such as PHLAT, HLA-VBseq, and SNP2HLA have been used in research settings, but performance heterogeneity among software tools is well documented. In the ultra-deep sequencing evaluation, the same input data produced different accuracy levels depending on which tool performed the analysis. This finding underscores the importance of validating any HLA typing tool against known samples before relying on it for research or clinical decisions.

## Sequencing Depth Requirements for HLA Variant Calling

Sequencing depth directly determines the accuracy of variant calling in the HLA region. An empirical evaluation using ultra-deep whole-genome sequencing data at approximately 410-fold coverage provided concrete depth thresholds for different analysis goals. The study randomly sampled sequence reads to construct simulation datasets at 54 different depths ranging from 0.05-fold to 410-fold, then evaluated genotype concordance against SNP microarray data and against the full-depth data.

For whole-genome variant calling, depths above 13.7-fold achieved greater than 99 percent concordance with SNP microarray genotypes. Single nucleotide variants reached greater than 95 percent concordance with the full-depth data at 17.6-fold coverage. Insertions and deletions were more demanding, showing only 60 percent concordance at that same depth. This finding indicates that accurate indel calling requires substantially higher coverage than SNV calling.

For HLA allele genotyping specifically, 13.7-fold depth showed sufficient accuracy, but the performance varied by software tool. Increasing depth beyond this level produced limited improvement in HLA genotyping accuracy. The practical implication is that a medium depth setting of approximately 15-fold achieves both accurate SNV calling and cost-effectiveness for whole-genome studies, while higher depths are needed when indels are the primary focus.

These depth recommendations apply to whole-genome sequencing. Whole-exome sequencing captures the coding regions of HLA genes through targeted enrichment, and the effective depth at HLA loci depends on the capture design and the efficiency of enrichment for this GC-rich and polymorphic region. Researchers should verify that their exome capture kit includes probes for the HLA genes of interest and should assess on-target depth at these loci specifically.

## Workflow Design for HLA-Aware Variant Calling

A practical workflow for variant calling in the HLA region combines standard whole-genome analysis with specialized HLA typing. The workflow should be designed from the start with the HLA region in mind, instead of treating HLA analysis as an afterthought.

The first step is data quality assessment. Raw sequencing reads should be evaluated for base quality, adapter contamination, and duplication rate. The HLA region's high GC content can cause coverage drops, so per-locus coverage in the HLA genes should be examined separately from genome-wide coverage statistics. The Galaxy Training Network provides accessible tutorials for quality control and read processing that can be adapted for this purpose.

The second step is alignment. For the genome-wide analysis, standard alignment to the reference genome is appropriate, but the alignment should be inspected for the HLA region. Reads with low mapping quality in the HLA genes should be expected, and their presence indicates that the specialized HLA analysis will be necessary. The alignment step should produce a BAM file with read group information, which is required for downstream variant calling and for the specialized HLA tools.

The third step is specialized HLA genotyping. The raw sequencing reads, not the aligned BAM file, should be used as input for tools like T1K or HLA-HD. These tools perform their own read processing and do not rely on the reference-based alignment that causes problems in the HLA region. Running the specialized tool on the same data that feeds the standard pipeline allows direct comparison of results.

The fourth step is integration of results. The genome-wide variant calls will include variants in the HLA region, but these should be treated with caution. The specialized HLA typing results provide the allele-level information that the standard pipeline cannot produce. When the two approaches disagree, the specialized HLA result should generally be trusted for allele assignment, but the discrepancy should be documented and investigated.

The fifth step is validation. For any study where HLA typing accuracy is critical, validation against a gold standard is essential. This validation can use samples with known HLA types from previous typing, samples typed by a different method such as sequence-specific primer PCR, or publicly available reference samples with established HLA types. The validation should assess both allele-level accuracy and variant-level concordance.

## Reproducible Workflow Implementation

Reproducibility is a core requirement for any bioinformatics analysis, and the HLA region's complexity makes reproducibility even more important. A reproducible workflow ensures that the same input data produces the same results regardless of who runs the analysis or when it is run.

The nf-core documentation describes community standards for pipeline development that emphasize reproducibility, portability, and version control. These standards include containerized software environments, pinned tool versions, and structured pipeline parameters. Adopting these standards for HLA analysis ensures that tool updates or environment changes do not silently alter results.

The Bioconductor project provides official documentation for reproducible genomic analysis workflows in R. For researchers who prefer R-based analysis, Bioconductor packages can handle the post-processing of HLA typing results, statistical analysis of allele associations, and visualization of results. The project's documentation emphasizes reproducible research practices, including versioned package management and literate programming.

The Carpentries lessons provide foundational training in shell, Git, and programming that supports reproducible analysis. Version control of analysis scripts and workflow definitions is essential for tracking how results were produced. A workflow that cannot be reproduced from raw data to final results has limited scientific value, particularly in a region as technically challenging as the HLA.

Containerization is a practical tool for reproducibility. Tools like T1K and HLA-HD have specific software dependencies, and these dependencies can change over time. Container images that bundle the tool with its exact dependencies ensure that the analysis runs identically across different computing environments. The nf-core documentation provides guidance on container usage within pipeline frameworks.

## At a Glance: HLA Variant Calling Decision Table

| Analysis Scenario | Recommended Approach | Expected Performance | Key Limitation |
| --- | --- | --- | --- |
| Whole-genome sequencing at 15x depth for population studies | Standard SNV calling plus HLA-HD or T1K for allele assignment | Greater than 99 percent concordance for SNVs, 96.9 percent HLA concordance with HLA-HD | Indel calling accuracy remains low at this depth |
| Whole-exome sequencing for disease association | Direct HLA allele calling with HLA-HD, not imputation | More accurate than array-based imputation, improved power for rare alleles and non-European ancestries | Exome capture must include HLA loci and achieve adequate on-target depth |
| RNA-seq for expression and allele analysis | T1K for joint HLA and KIR genotyping | Accurate allele inference from transcriptome data, can process single-cell data | Expression level affects coverage and may bias allele detection |
| Ultra-deep sequencing for clinical-grade typing | High-depth whole-genome or targeted HLA sequencing with validated tool | Highest accuracy, but improvement plateaus beyond approximately 15x for HLA typing | Increased cost and the need for tool-specific validation |

## Data Inputs and Their Impact on HLA Analysis

The choice of sequencing data type fundamentally shapes what HLA analysis is possible. Whole-genome sequencing provides uniform coverage across the HLA region, including introns and regulatory regions, but the cost is higher than other approaches. Whole-exome sequencing targets coding regions and reduces cost, but the capture design determines whether HLA genes are adequately represented. RNA-seq provides information about expressed alleles and can reveal allele-specific expression, but the analysis depends on the transcriptome's composition and the genes expressed in the sampled tissue.

Whole-genome sequencing at approximately 15-fold depth provides a practical balance for studies that need both genome-wide variant calling and HLA typing. The empirical evaluation of ultra-deep sequencing data showed that this depth achieves high SNV concordance and sufficient HLA genotyping accuracy. Studies focused on indels should plan for higher depth, since indel concordance at 17.6-fold was only 60 percent.

Whole-exome sequencing from large biobanks has proven valuable for HLA disease association studies. The UK Biobank application of HLA-HD to 454,824 participants demonstrated that direct HLA allele calling from exome data is feasible at scale and more accurate than imputation. This approach identified 360 associations for 11 autoimmune phenotypes, with at least 129 likely novel. The study also showed that HLA alleles with synonymous variants, which are often overlooked, can significantly influence phenotypes.

RNA-seq data enables HLA and KIR genotyping from the transcriptome. T1K was specifically designed to work with RNA-seq data and can process single-cell RNA-seq data. This capability opens opportunities for studying HLA and KIR expression in specific cell populations, such as the finding that KIR2DL4 expression was enriched in tumor-specific CD8+ T cells. Researchers using RNA-seq for HLA analysis should be aware that genes with low or absent expression will not be typed, and the results reflect the expressed alleles instead of the genomic genotype.

## Options and Tradeoffs in HLA Typing Tools

The choice of HLA typing tool involves tradeoffs among accuracy, speed, input data requirements, and the ability to handle specific gene families. No single tool performs best in all scenarios, and the empirical evidence shows performance heterogeneity among tools on the same data.

HLA-HD has demonstrated high accuracy in multiple evaluations and has been applied at biobank scale. Its strengths include accurate allele calling from whole-genome and whole-exome data and the ability to handle the HLA region's complexity. The UK Biobank study showed that direct calling with HLA-HD outperforms imputation, particularly for rare alleles and non-European ancestries. The main limitation is that HLA-HD focuses on HLA genes and does not address KIR genes.

T1K provides the unique capability of joint HLA and KIR genotyping from the same data. This joint modeling is essential for KIR genes because many KIR genes are similar to each other and may be absent from some chromosomes. T1K's ability to distinguish homologous genes like KIR2DL5A and KIR2DL5B addresses a challenge that other tools cannot solve. T1K also supports RNA-seq and single-cell data, making it suitable for expression studies. The tradeoff is that T1K may require more computational resources than simpler tools, and its performance should be validated for the specific data type being used.

Imputation-based approaches, such as SNP2HLA, infer HLA alleles from genotyping array data using reference panels. These approaches are cost-effective because they use existing array data, but they have lower accuracy than direct sequencing-based calling. The UK Biobank study demonstrated that imputation decreases accuracy and statistical power for rare alleles and in non-European ancestries. Imputation remains useful for studies where only array data are available, but direct calling is preferred when sequencing data exist.

The performance heterogeneity among tools means that validation is essential. A tool that performs well on whole-genome data may perform differently on exome data or RNA-seq data. Researchers should validate their chosen tool against samples with known HLA types before committing to a large-scale analysis.

## Practical Implementation Steps for HLA Variant Calling

Implementing HLA variant calling in a research or clinical laboratory requires a structured approach that addresses the region's unique challenges. The following steps provide a practical framework for implementation.

The first step is to define the analysis goal. Is the study investigating HLA allele associations with disease, characterizing HLA expression, or integrating HLA variants into a broader genome-wide analysis? The goal determines the required resolution, the appropriate data type, and the necessary validation.

The second step is to select the sequencing strategy. Whole-genome sequencing at approximately 15-fold depth supports both genome-wide variant calling and HLA typing. Whole-exome sequencing is appropriate when the capture design includes HLA loci and when the study population is large. RNA-seq is appropriate for expression studies. The sequencing strategy should be finalized before data collection begins.

The third step is to establish the analysis environment. This includes installing the chosen HLA typing tools, setting up containerized environments for reproducibility, and documenting the exact software versions and parameters. The nf-core documentation and Bioconductor project provide guidance on reproducible analysis environments.

The fourth step is to run the analysis on a validation set. Samples with known HLA types should be analyzed first to assess the tool's accuracy on the specific data type and depth being used. The validation results should be documented and compared against the expected types. If the validation fails, the tool choice or sequencing strategy should be reconsidered.

The fifth step is to run the full analysis with the validated pipeline. The pipeline should include quality control at each stage, from raw reads through alignment to final variant calls. The HLA-specific results should be integrated with the genome-wide results, and discrepancies should be investigated.

The sixth step is to document the analysis and results. The documentation should include the pipeline version, tool parameters, validation results, and any deviations from the standard workflow. This documentation supports reproducibility and provides the context needed for interpreting results.

## Records and Measurements for HLA Analysis

Maintaining detailed records is essential for HLA variant calling because the analysis involves multiple tools, parameters, and quality metrics. The following records should be maintained for each sample or study.

The raw sequencing metrics should include total reads, read length, base quality scores, and duplication rate. For whole-genome sequencing, the mean depth and the depth distribution across the genome should be recorded. For the HLA region specifically, per-locus depth should be recorded for each HLA gene, since the region's GC content can cause coverage variation.

The alignment metrics should include the overall alignment rate, the percentage of reads with mapping quality above the threshold used for variant calling, and the specific alignment statistics for the HLA region. Low mapping quality in the HLA region is expected and should be documented as part of the analysis.

The HLA typing results should include the assigned alleles at each locus, the confidence scores or quality metrics provided by the tool, and the version of the HLA allele database used for assignment. HLA allele nomenclature changes over time, so the database version is essential for interpreting results.

The validation records should include the known HLA types for validation samples, the called types, and the concordance calculation. Any discordant calls should be investigated and documented. The validation results provide the evidence needed to assess the reliability of the full analysis.

The computational records should include the software versions, container images, and parameters used for each analysis step. These records support reproducibility and troubleshooting. The nf-core documentation provides standards for documenting pipeline usage and configuration.

## Common Failure Patterns in HLA Variant Calling

Recognizing common failure patterns helps researchers identify problems early and avoid incorrect conclusions. The following patterns are frequently observed in HLA variant calling.

The first failure pattern is low or zero coverage at specific HLA loci. This occurs when the sequencing library preparation or capture design does not adequately represent the HLA genes, or when the alignment algorithm cannot place reads due to polymorphism. The result is missing genotype information at affected loci. This pattern is detected by examining per-locus coverage in the HLA region.

The second failure pattern is an excess of apparent variants in the HLA region. When reads from multiple similar HLA genes align to a single locus, the variant caller may report many false variants that reflect the differences between the genes instead of true variation in the sample. This pattern is detected by comparing the variant density in the HLA region to the genome-wide average and by examining the allele frequencies of the reported variants.

The third failure pattern is discordance between the genome-wide variant calls and the specialized HLA typing results. The genome-wide pipeline may report variants that the HLA typing tool does not support, or the HLA typing tool may assign an allele that is inconsistent with the genome-wide variants. This discordance indicates that one or both analyses have errors, and the discrepancy should be investigated before drawing conclusions.

The fourth failure pattern is poor performance on validation samples. If the chosen tool does not accurately type samples with known HLA types, the tool or the sequencing data may be inadequate. The validation results should be examined to determine whether the errors are concentrated at specific loci, specific allele groups, or specific data types.

The fifth failure pattern is batch effects in large studies. When samples are sequenced in multiple batches, differences in library preparation, sequencing platform, or capture design can introduce systematic variation in HLA typing accuracy. The UK Biobank study's success with a large cohort demonstrates that batch effects can be managed, but they require explicit attention in the analysis design.

## Limitations of HLA Variant Calling Approaches

Every approach to HLA variant calling has limitations that should be understood before interpreting results. The most fundamental limitation is that short-read sequencing cannot fully resolve the HLA region's complexity. Reads of 100 to 150 base pairs are shorter than the polymorphic segments that distinguish many HLA alleles, and the alignment ambiguity cannot be completely eliminated regardless of the analysis method.

The resolution of HLA typing is another limitation. Many tools report alleles at the field level, such as HLA-A*02:01, but do not resolve all nucleotide differences that distinguish closely related alleles. Higher resolution typing may require long-read sequencing or targeted approaches that are not part of standard variant calling workflows.

The reference bias in HLA analysis is a persistent limitation. Tools that align to a reference genome are biased toward detecting variants that differ from the reference allele, and this bias can affect allele frequency estimates and association studies. The UK Biobank study's finding that direct calling outperforms imputation, particularly for non-European ancestries, reflects the reference bias in imputation panels.

The interpretation of synonymous variants in HLA genes is an emerging area with limitations. The UK Biobank study showed that synonymous variants can influence autoimmune phenotypes, but the mechanisms are not fully understood. Researchers should not assume that synonymous variants are neutral in the HLA region.

The population-specific nature of HLA variation limits the generalizability of findings. The recurrent pregnancy loss study identified HLA-C*12:02, HLA-B*52:01, and HLA-DRB1*15:02 as protective alleles in a Japanese population, but these alleles may have different effects in other populations. The long-range haplotype structure in the HLA region means that associations may reflect linked variants instead of the typed allele itself.

## Quality Controls for HLA Variant Calling

Quality control is essential at every stage of HLA variant calling, and the controls should be tailored to the region's specific challenges. The following controls should be implemented in any HLA analysis workflow.

Sample-level quality control should verify that the sample identity is correct and that the sequencing data meet minimum quality standards. This includes checking the concordance between reported sex and the sex chromosomes, verifying that the sample's genetic ancestry is consistent with the study design, and confirming that the sequencing depth meets the study's requirements.

Read-level quality control should assess base quality scores, adapter contamination, and duplication rates. The HLA region's high GC content can cause specific quality issues, so the quality metrics should be examined for the HLA loci separately from the genome-wide metrics.

Alignment-level quality control should examine the mapping quality distribution, the coverage uniformity across the HLA region, and the presence of alignment artifacts. Low mapping quality in the HLA region is expected, but the pattern of low mapping quality should be documented and compared across samples.

Variant-level quality control should assess the variant quality scores, the allele balance, and the concordance between the genome-wide and HLA-specific analyses. Variants in the HLA region with unusual allele balance or quality scores should be flagged for review.

HLA typing quality control should include the confidence scores provided by the typing tool, the consistency of results across loci, and the concordance with any available validation data. The HLA allele database version should be recorded, and the results should be interpreted in the context of that version.

## Safety and Regulatory Context for HLA Analysis

HLA analysis has clinical and regulatory implications that researchers and laboratory professionals must consider. The HLA region's role in transplantation matching means that HLA typing results can influence clinical decisions, and the accuracy requirements for clinical use are higher than for research use.

Laboratories performing HLA typing for clinical purposes should follow the applicable regulatory standards for their jurisdiction. These standards typically require validation of the typing method, participation in proficiency testing, and documentation of the laboratory's procedures. The specific requirements vary by jurisdiction and by the intended use of the results.

The interpretation of HLA typing results should be performed by qualified personnel who understand the region's complexity and the limitations of the analysis method. Automated variant calling and HLA typing tools provide results, but the interpretation of those results in a clinical context requires expertise.

The reporting of HLA typing results should include the method used, the resolution achieved, and any limitations of the analysis. The HLA allele nomenclature should follow the current World Health Organization HLA Nomenclature Committee standards, and the database version used for allele assignment should be reported.

The use of HLA typing results in disease association studies has ethical considerations. The HLA region is associated with many diseases, and the results of HLA analysis may have implications for research participants. Studies should follow the ethical standards for genetic research, including informed consent and data protection.

## Professional Escalation Criteria for HLA Analysis

Researchers and laboratory professionals should know when to escalate HLA analysis issues to specialists. The following situations warrant escalation to a bioinformatics specialist, a clinical immunogeneticist, or a laboratory director.

Discordance between genome-wide variant calls and specialized HLA typing results should be escalated when the discrepancy affects the study's conclusions. The investigation of the discordance may require expertise in both the analysis methods and the HLA region's biology.

Validation failures should be escalated when the chosen tool does not accurately type validation samples. The failure may indicate a problem with the tool, the sequencing data, or the validation design, and the resolution may require specialized expertise.

Unexpected HLA typing results should be escalated when they have clinical implications or when they conflict with established knowledge about the sample or the study population. The HLA region's complexity means that unexpected results may reflect true biology or technical artifacts, and distinguishing between the two requires expertise.

The need for higher resolution HLA typing should be escalated when the current analysis cannot resolve alleles that are clinically or scientifically important. The decision to use long-read sequencing or other advanced methods should be made with input from specialists who understand the options and their limitations.

The interpretation of HLA associations in disease studies should be escalated when the results have implications for clinical care or when the study's conclusions depend on the accuracy of the HLA typing. The HLA region's linkage disequilibrium and population-specific variation require careful interpretation.

## Frequently Asked Questions

### Why do standard variant calling pipelines fail in the HLA region?

Standard pipelines align short reads to a reference genome and call variants where the reads differ from the reference. The HLA region contains multiple highly similar genes, so reads from one HLA gene can align to several genes or alleles. Aligners assign low mapping quality to these reads, and variant callers filter them out, creating coverage gaps and missed variants. The reference genome also contains one specific HLA allele at each locus, so reads from individuals with different alleles have more mismatches and may be placed suboptimally. The region's structural variation and copy number changes further violate the assumptions of standard diploid variant callers.

### What is the difference between HLA genotyping and standard variant calling?

Standard variant calling identifies individual nucleotide changes relative to a reference genome. HLA genotyping assigns specific allele names to each HLA gene based on the combination of variants present. Because HLA alleles are defined by their full sequence, researchers need to know which variants occur together on the same chromosome. Standard variant callers do not provide this phased allele information reliably in the HLA region. Specialized HLA typing tools assemble the variants into allele-level haplotypes and report the assigned alleles.

### Which HLA typing tool should I use for whole-genome sequencing data?

The choice depends on your specific needs. HLA-HD has demonstrated high accuracy in multiple evaluations, achieving 96.9 percent concordance in an ultra-deep sequencing study, and has been applied successfully to whole-exome data from 454,824 UK Biobank participants. T1K provides the additional capability of joint HLA and KIR genotyping and supports RNA-seq and single-cell data. You should validate any tool against samples with known HLA types before relying on it, since performance heterogeneity among tools is well documented.

### What sequencing depth is needed for accurate HLA variant calling?

Whole-genome sequencing at approximately 15-fold depth provides sufficient accuracy for both genome-wide SNV calling and HLA allele genotyping. An empirical evaluation using ultra-deep sequencing data showed that depths above 13.7-fold achieved greater than 99 percent concordance with SNP microarray genotypes, and 13.7-fold depth showed sufficient accuracy for HLA genotyping. Indel calling requires higher depth, with only 60 percent concordance at 17.6-fold depth. Increasing depth beyond approximately 15-fold produces limited improvement in HLA genotyping accuracy.

### Can I use whole-exome sequencing for HLA analysis?

Yes, whole-exome sequencing can be used for HLA analysis if the capture design includes the HLA genes and achieves adequate on-target depth. The UK Biobank study successfully called HLA alleles from whole-exome sequencing data from 454,824 participants using HLA-HD and found that direct calling was more accurate than imputation from genotyping arrays. You should verify that your exome capture kit includes probes for the HLA genes of interest and assess the on-target depth at these loci specifically.

### How does imputation compare to direct HLA allele calling?

Imputation infers HLA alleles from genotyping array data using reference panels, while direct calling determines alleles from sequencing data. Direct calling is more accurate, particularly for rare alleles and in non-European ancestries. The UK Biobank study demonstrated that imputation decreases accuracy and statistical power, and that direct calling from whole-exome sequencing identified more associations with autoimmune phenotypes. Imputation remains useful when only array data are available, but direct calling is preferred when sequencing data exist.

### What are the limitations of short-read sequencing for HLA analysis?

Short reads of 100 to 150 base pairs cannot fully resolve the HLA region's complexity. The reads are shorter than the polymorphic segments that distinguish many HLA alleles, so alignment ambiguity cannot be completely eliminated. The reference bias in short-read analysis systematically affects the detection of variants that differ from the reference allele. Higher resolution typing may require long-read sequencing or targeted approaches. The HLA allele database version also affects results, since allele nomenclature changes over time.

### How should I validate my HLA variant calling pipeline?

Validation should use samples with known HLA types from a different method, such as sequence-specific primer PCR, or publicly available reference samples with established types. Run the validation samples through your complete pipeline, from raw reads to final allele calls, and compare the results to the known types. Document the concordance and investigate any discordant calls. The validation should be repeated when you change the sequencing platform, the typing tool, the tool version, or the analysis parameters.

## Related Bioinformatics Guides

- [Detecting Structural Variants with Long-Read Sequencing: Methods and Considerations](/knowledge/bioinformatics/detecting-structural-variants-with-long-read-sequencing-methods-and-considerations)
- [Variant Calling GATK: Structural Analysis and Computational Methodologies in Bioinformatics](/knowledge/bioinformatics/variant-calling-gatk)
- [Variant Calling Pipelines: GATK Best Practices, FreeBayes, and DeepVariant Comparison](/knowledge/bioinformatics/variant-calling-pipelines-gatk-deepvariant)
- [Oxford Nanopore Sequencing: From Sample to Base Calls](/knowledge/bioinformatics/oxford-nanopore-sequencing-from-sample-to-base-calls)
- [Long-Read Metagenome Assembly: Overcoming Challenges with Nanopore and PacBio Data](/knowledge/bioinformatics/long-read-metagenome-assembly-overcoming-challenges-with-nanopore-and-pacbio-data)

## Related Clinical & Scientific Guides

* [A Practical Guide to Detecting Antimicrobial Resistance Genes in Shotgun Metagenomic Data](/knowledge/bioinformatics/a-practical-guide-to-detecting-antimicrobial-resistance-genes-in-shotgun-metagenomic-data)
* [Computational Immunology: Modeling the Immune System](/knowledge/bioinformatics/computational-immunology-modeling-the-immune-system)
* [How to Set Hard Filters for Germline Variant Calling: A Practical Guide to GATK Best Practices](/knowledge/bioinformatics/how-to-set-hard-filters-for-germline-variant-calling-a-practical-guide-to-gatk-best-practices)


## References and Further Reading

- [NCBI Data Resources](https://www.ncbi.nlm.nih.gov/). National Center for Biotechnology Information.
- [EMBL-EBI Training](https://www.ebi.ac.uk/training). European Bioinformatics Institute.
- [Bioconductor](https://bioconductor.org/). Bioconductor Project.
- [Galaxy Training Network](https://training.galaxyproject.org/). Galaxy Project.
- [nf-core Documentation](https://nf-co.re/docs). nf-core.
- [The Carpentries Lessons](https://carpentries.org/lessons). The Carpentries.
- [Efficient and accurate KIR and HLA genotyping with massively parallel sequencing data.](https://pubmed.ncbi.nlm.nih.gov/37169596). Genome research, 2023.
- [HLA class I signal peptide polymorphism determines the level of CD94/NKG2-HLA-E-mediated regulation of effector cell responses.](https://pubmed.ncbi.nlm.nih.gov/37264229). Nature immunology, 2023.
- [Common and rare genetic variants predisposing females to unexplained recurrent pregnancy loss.](https://pubmed.ncbi.nlm.nih.gov/39019884). Nature communications, 2024.
- [HLA allele-calling using multi-ancestry whole-exome sequencing from the UK Biobank identifies 129 novel associations in 11 autoimmune diseases.](https://pubmed.ncbi.nlm.nih.gov/37923823). Communications biology, 2023.
- [Empirical evaluation of variant calling accuracy using ultra-deep whole-genome sequencing data.](https://pubmed.ncbi.nlm.nih.gov/30741997). Scientific reports, 2019.

> This article is educational and does not replace validated analysis plans, institutional policy, clinical interpretation, or specialist review.