BACKGROUND: Human Leukocyte Antigens (HLA) play central roles in histocompatibility and immune system functions, including antigen presentation. Accurate typing of Class I and II HLA genes is crucial for transplant tissue matching, characterising autoimmune diseases and informing cancer immunotherapy. Clinical serology and PCR-based testing are the gold standards for HLA typing, but offer only single-field resolution (e.g., HLA-A*11). Whole genome sequencing (WGS) and RNA sequencing (RNA-seq) can achieve higher, three-field resolution (e.g., HLA-A∗11:01:01), although some HLA genes can be challenging to type from sequencing data. With the increasing use of germline WGS, tumour WGS and tumour RNA-seq in cancer patient care, there is an opportunity to combine these three dataset types to improve HLA typing accuracy and confidence, and to identify clinically relevant HLA type changes in tumours. To achieve this, we developed consHLA, a tool that employs this consensus HLA typing approach. RESULTS: We obtained matched germline and tumour WGS and RNA-seq data from 86 high-risk paediatric cancer patients (76 brain cancers, 10 leukaemias) from the ZERO Childhood Cancer precision medicine program. We examined 10 HLA typing packages, selecting HLA-HD to develop our consHLA workflow as HLA-HD can employ all three dataset types, analysing both Class I and II HLA genes at three field resolution. Using consHLA we achieved 97.9% concordance with gold standard HLA test results. We observed 90.5% allele consistency across the three sequencing NGS inputs. Typing inconsistencies in at least one of 12 clinically relevant HLA genes were observed in 29 of the brain tumour cases. 32% of these had clinically relevant explanations. To assist clinically, we implemented consHLA as a fully automated workflow producing a clinician-friendly HLA-typing report. CONCLUSIONS: To leverage cancer patient germline and tumour WGS and tumour RNA-seq data we developed an automated workflow, consHLA, that produces consensus typing of HLA genes in a clinically relevant timeframe. This workflow provides higher resolution patient HLA-typing than current gold standard approaches, identifies HLA alterations arising in patient tumours and generates clear, simple reports.

Motivation: Genome sequencing has had a remarkable impact on our ability to study the effects of human genetic variation, however, variant interpretation remains the major bottleneck. Understanding the potential impact of variants, including structural variants, requires extensive annotation from disparate sources of knowledge, and in silico prediction algorithms. Results: We introduce Seave, an intuitive web platform that enables all types of variants to be securely stored, annotated and filtered. Variants are annotated with allele frequencies and pathogenicity assessments from many popular databases and in silico pathogenicity prediction scores. Seave enables filtering of variants with specific inheritance patterns, including somatic variants, by quality, allele frequencies and gene lists which can be curated and saved. Seave was made for whole genome data and is capable of storing and querying copy number and structural variants. Availability and implementation: To demo Seave with public data, see https://www.seave.bio. Source code is available at http://code.seave.bio and extensive documentation is available at http://documentation.seave.bio. Seave can be locally installed on an Apache server with PHP and MySQL, or we provide an Amazon Machine Image for quick deployment. For commercial and clinical diagnostic licensing, contact the corresponding author. Supplementary information: Supplementary data are available at Bioinformatics online.