gms | German Medical Science

Introduction: The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic continues to be a health threat. Constantly emerging mutations in the SARS-CoV-2 genome enable effective evasion of available treatments. It is assumed that some infected patients carry multiple versions of the virus, a so-called majority variant (or consensus sequence) and one or more minority variants, representing additional mutations. These minority variants potentially lead to greater immune evasion capacities and higher rates of transmissibility [1].

Even though the effect of minority variants towards treatment methods is known and the Robert Koch institute stated the need to identify these variants automatically and accurately within SARS-CoV-2 sequence data in patient samples, a high-fidelity analysis for these variants does not currently exist [2]. We developed a first algorithm towards a high-fidelity analysis method to detect SARS-CoV-2 minority variants within sequencing data of patient samples. The analysis algorithm is set out to identify all SARS-CoV-2 minority variants and mutations. It provides information about the mutation frequencies and their positions as well as the current classification and pangolin lineage of the respective sample. The algorithm is an extension to the UnCoVar workflow by Thomas and Battenfeld et al. (unpublished) [3], a transparent and robust workflow to analyze SARS-CoV-2 sequencing data.

In contrast to other bioinformatic tools used to identify new SARS-CoV-2 variants, which rely on consensus sequences considering only majority events, the algorithm considers all mutations independent of their frequency. This approach enables the detection of mutations with a higher resolution and provides a research platform to analyze single nucleotide variants (SNVs) and possible mutation patterns within SARS-CoV-2 sequencing sets.

The algorithm will provide a platform for high-fidelity analyses of all detectable mutations and variants, which is crucial for detection, monitoring and description of changes within the SARS-CoV-2 genome. This is essential since treatment thrives when adjusted to patient specific mutations patterns, including cases of infections with one or more variants.

Methods: The analysis conducted by the developed algorithm is based on the UnCoVar workflow by Thomas and Battenfeld et al. (unpublished), using Snakemake [3]. The UnCoVar workflow is embedded in a continuous integration and continuous delivery (CI/CD) environment. This type of production environment applies an ongoing automation, testing and monitoring to improve the development process, producing reliable reproducible data.

Sequencing data (e.g., produced by the Illumina platform) is provided to the UnCoVar workflow in a paired-end FASTQ format, which excludes reads shorter than 30 base pairs and a Phred quality below 20. Pre-processing and quality control is conducted using fastp [4] bwa [5] and BAMclipper [6]. Artificially introduced sequencing adapters are trimmed and non-viral sequencing reads are removed based on an alignment against a combined reference of the human (GRCh38.p13) and the SARS-CoV-2 reference genome (NC_045512.2). Afterwards, the raw reads are assembled to contigs using MEGAHIT [7] and the resulting contigs are scaffolded against the SARS-CoV-2 reference genome using raGOO [8]. Freebayes [9] and vep [10] are used to detect single and multiple nucleotide variants (SNVs, MNVs), small insertions and deletions (indels) within the scaffolds and structural variants are identified using Delly [11]. These variants are evaluated based on a unified statistical model using Varlociraptor [12]. Afterwards, pangolin [13] is used to employ the dynamic nomenclature for SARS-CoV-2 lineages, suggested by Rambaut et al. (2020) [14].

The output of the UnCoVar workflow provides the base for the newly developed algorithm to identify and extract minority mutations in SARS-CoV-2 qPCR-positive patient samples of University Hospital of Essen, Germany, covering a period from February 2021 to October 2022. The python-based algorithm extracts the read depth, frequency and position of the mutations as well as the pangolin lineage and its classification of the reconstructed genomes from the UnCoVar output and converts the data to be accessible for further analysis of minority variants. Afterwards, the data can be structured and filtered for investigation of specific timeframes, samples, mutations or similar use cases.

Results: We developed and implemented an algorithm, which reliably filters different minority mutations and variants in a fast and reliable high throughput manner. Out of 2,188 samples, with 357,847,205,526 base pairs a total of 744,657 mutations containing 50,050 instances of minority variants were detected. First implementations of the algorithm to study low frequent mutation patterns within patient sample data identified a rare mutation (S:D820A), which evades a group of recently identified fusion peptide-binding broadly neutralizing antibodies (FP-bnAb) [15], [16].

Outlook: The algorithm or resulting tool to detect and study SARS-CoV-2 minority events needs to be benchmarked further and integrated into a continuous integration and continuous delivery (CI/CD) environment. This type of production environment applies an ongoing automation, testing and monitoring to improve the development process, resulting in reliable and functioning software, producing reliable reproducible data.

The implemented algorithm will provide a platform to search for newly emerging SARS-CoV-2 variants, which can evade currently applied treatment methods, in an accurate and high-throughput manner. This includes further investigation of recombination events and thereby enables a more thorough monitoring of emerging Sars-CoV-2 variants.