Addressing data management and analysis challenges in viral genomics: The Swiss HIV cohort study viral next generation sequencing database

Abstract

Numerous HIV related outcomes can be determined on the viral genome, for example, resistance associated mutations, population transmission dynamics, viral heritability traits, or time since infection. Viral sequences of people with HIV (PWH) are therefore essential for therapeutic and research purposes. While in the first three decades of the HIV pandemic viral genomes were mainly sequenced using Sanger sequencing, the last decade has seen a shift towards next-generation sequencing (NGS) as the preferred method. NGS can achieve near full length genome sequence coverage and simultaneously, it accurately encapsulates the within-host diversity by characterizing HIV subpopulations. NGS opens new avenues for HIV research, but it also presents challenges concerning data management and analysis. We therefore set up the Swiss HIV Cohort Study Viral NGS Database (SHCND) to address key issues in the handling of NGS data including high loads of raw- and processed NGS data, data storage solutions, downstream application of sophisticated bioinformatic tools, high-performance computing resources, and reproducibility. The database is nested within the Swiss HIV Cohort Study (SHCS) and the Zurich Primary HIV Infection Cohort Study (ZPHI), which together enrolled 21,876 PWH since 1988 and include a biobank dating back to the early nineties. Since its initiation in 2018, the SHCND accumulated NGS sequences (plasma and proviral origin) of 5,178 unique PWH. We here describe the design, set-up, and use of this NGS database. Overall, the SHCND has contributed to several research projects on HIV pathogenesis, treatment, drug resistance, and molecular epidemiology, and has thereby become a central part of HIV-genomics research in Switzerland.

Fig 1. From sample to genome.

Description of the steps from blood sampling of people with HIV, Next Generation Sequencing and its data handling, to the bio-informatics tools for computational HIV genome analysis.

Fig 2. Illustration of the database storage.

(A) NGS storage and (B) Bioinformatic result storage implementation in SHCND. Basically, a graph of UUID labeled nodes is formed. The graph structure (edges) itself is also materialized in UUID-named files containing JSON (not shown). Furthermore, the exact processing tool used to generate each result container is included in the metadata of each result container.

Fig 3. Illustration of the bioinformatic pipeline processing.

From initial job submission to storage of processed results. In the case of “global” processing tools, multiple samples’ fastq and output files from previous processing tool runs can be retrieved as inputs to the Processing step (not shown).

Fig 4. Sequence availability and timing.

(A) Sequences available per gene and for the whole genome in the Swiss HIV Cohort Study Viral NGS Database (SHCND), stratified by sample source. Viral gene sequences considered, cover at least 40% of the respective gene length compared to the HIV-1 HXB2 gene reference sequence (GenBank accession number K03455). (B) Cumulative sample counts by sampling year and year when NGS was performed, stratified by HIV-1 plasma RNA and HIV-1 proviral DNA.