Table 1.

Details of the five HLA typing algorithms included in this project.

Tool	Version	Resolution	Approach	Reason for inclusion in this study	Known disadvantages
Kourami (21)	0.9.6	G group	Weighted graph structure from alignment of input reads aligned to reference sequences. Most probable graph path is the inferred type.	High typing accuracy in previous studies (39). Can detect and report new alleles, which are not included in a database.	Build for WGS based and is negatively affected by gaps in the sequencing data.
HLA*LA (23)	1.0.1	G group	Linear alignments projected on a population reference graph. Likelihood functions to infer the HLA type.	High typing accuracy in previous studies (39).
HISAT-genotype (20)	1.3.2	4-field	Graph-based alignment (HISAT2) and an expectation maximisation algorithm.	High typing accuracy in previous studies (40). Able to detect novel alleles. Unique, as it does not include some form of linear alignment, but an extension of BWT for graphs.
STC-Seq (22)	1.0	3-field	Dense chip-based probes that capture the coding regions of HLA. Linear alignment algorithm.	Offers a perspective by showing the performance of a simpler bioinformatics approach, as it is designed for HLA enriched data.	Not designed for general sequencing data.
Optitype (19)	1.3.3	2-field	Integer linear programming to find the allele combination that explains the highest number of reads.	High typing accuracy in previous studies (25, 26, 28)

Each of the original articles describing the tools contains some sort of benchmarking study, demonstrating the capabilities of the tool. For more a more extensive outline of these five HLA typing tools, see Supplementary Table S1 .