FuFiHLA: Full Field HLA allele typing for Long Reads

FuFiHLA is a pipeline for full-field HLA allele typing and consensus sequence construction from long-read sequencing data. It performs complete allele typing and can additionally report partial (CDS-supported) alleles when mutations are detected within coding exons.

It currently supports PacBio HiFi data on six clinically important transplant genes: HLA-A, -B, -C, -DQA1, -DQB1, -DRB1.

Highlights

• Reference-free: does not depend on a specific version of reference genome such GRCh38 or CHM13
• Improved consensus accuracy compared to StarPhase

Citation: TBD

---

Installation

Install from Bioconda (recommended):

conda install -c bioconda -c conda-forge fufihla

---

Quick Test

With test.fa.gz under the folder test, run:

fufihla --fa test.fa.gz --out test_dir

The output includes:

• test_dir/ → pipeline logs
• test_dir.out → result output
• test_dir.err → stderr log

---

Usage

To use the latest reference allele sequences from IMGT, type:

fufihla-ref-prep

This will create a directory called ref_data, which would contain the reference allele sequence ref.gene.fa.gz.

Run the pipeline:

# default bundled reference (IPD-IMGT/HLA)
fufihla --fa <input_reads.fa.gz> --out <output_dir>

# specify reference directory and options
fufihla --fa <input_reads.fa.gz> --out <output_dir> --refdir <reference data directory> --hifi/--ont --hom-threshold <float> --debug

Arguments

• <input_reads.fa.gz> : raw long reads (.fa/.fa.gz/.fq/.fq.gz)
• <output_dir> : directory for pipeline outputs
• --refdir <reference_data_directory> (optional): path to reference allele dataset; if omitted, uses the default bundled set
• --hifi/--ont (optional): choose HiFi or Nanopore input reads, default is --hifi
• --hom-threshold <float> (optional): threshold for homozygous detection. When two haplotypes are highly similar, FuFiHLA suppresses a second allele call to avoid false heterozygous typing.
• --debug (optional): keep all intermediate files; otherwise only consensus results are retained

---

Outputs

A typical run produces:

<outdir>/consensus/*_asm*.fa

Consensus allele FASTA sequences for each gene haplotype.

Allele calls are printed to <output_dir>.out in PAF-like format with minimap2 tags.

Example:

HLA-A*01:01:01:01  cons_HLA-A*01_01_01_01  ...  cs:Z::3503
HLA-A*26:01:01:01  cons_HLA-A*26_01_01_01  ...  cs:Z::3517

• Column 1 → final allele call
• Column 2 → consensus sequence identifier
• Last column (cs:Z) → minimap2 cs tag describing base-level alignment

Interpretation:

• Known alleles: cs:Z::3503 → perfect match across the gene
• Novel alleles: substitutions *, insertions +, or deletions - appear in cs:Z

When partial alleles are reported, the allele originates from CDS alignment rather than full gene alignment. This occurs only when exon mutations prevent confident full-length allele assignment.

---

Running Tips

Extract reads from existing BAM files can also generate similar results:

echo "
chr6	29942254	29945755
chr6	31268254	31272571
chr6	31353362	31357442
chr6	32578769	32589848
chr6	32636717	32643200
chr6	32660031	32667132" > sel.bed

samtools view -bh ${bam} --region-file sel.bed | samtools fasta | gzip -c > out.fa.gz