HLA predictions from long sequence read alignments, streamed directly into HLAminer

TL;DR

This paper introduces a streamlined protocol for predicting HLA genotypes directly from long sequencing reads by streaming alignments into HLAminer, enabling efficient analysis of polymorphic human genome regions.

Contribution

The method simplifies HLA prediction from long reads by streaming alignments into HLAminer without storing large files, compatible with any sam-format aligner, and effective with lower-accuracy datasets.

Findings

Robust HLA predictions with older nanopore datasets

Effective at low 10X coverage

Compatible with various long read aligners

Abstract

The rapidly changing landscape of sequencing technologies brings new opportunities to genomics research. Longer sequence reads and higher sequence throughput coupled with ever-improving base accuracy and decreasing per-base cost is now making long reads suitable for analyzing polymorphic regions of the human genome, such as those of the human leucocyte antigen (HLA) gene complex. Here I present a simple protocol for predicting HLA signatures from whole genome shotgun (WGS) long sequencing reads, by directly streaming sequence alignments into HLAminer. The method is as simple as running minimap2, it scales with the number of sequences to align, and can be used with any read aligner capable of sam format output without the need to store bulky alignment files to disk. I show how the predictions are robust even with older and less [base] accurate WGS nanopore datasets and relatively low (10X) sequence coverage and present a step-by-step protocol to predict HLA class I and II genes from the long sequencing reads of modern third-generation technologies.