Temporally representative sequences to unravel the SARS-CoV-2 data complexity – Applications of the SARS-CoV-2 Mutation Frequency Calculator

With over 16 million international and over 1.6 million national full-sequencegenomes, SARS-CoV-2 is one of the most sequenced genomes. This amountof data and the associated sequencing errors and biases such as amplicondropouts pose a challenge for analysing SARS-CoV-2 data. Mutation frequenciesare a well-known and widely used medium in the scientific community,but there is no standardised method for this. Using the SARS-CoV-2 Mutation Frequency Calculator (SC2MFC), together with the foundationcovSonar, mutation frequencies of any time period can be calculated automaticallyand reproducibly. To overcome the challenges of large amounts ofdata and sequencing errors, temporally lineage identifiers and representativesequences are determined, making it possible to analyse data from wastewaterand build initial phylogenetic trees for large-scale analysis. In addition,the SC2MFC is used to validate previously arbitrary parameters based onstatistical tests such as a random bootstrap procedure and the calculation ofprecision recall based on machine learning. Results from the application ofSC2MFC on three previously defined data sets have shown, among otherthings, that a definition of the frequency threshold, such as this 75% usedby outbreak.info, cannot be used as a standard, but must be recalculated foreach period. A generalised minimum number of samples of a lineage cannotbe determined either, but is also dependent on pangolin lineages fromthe respective period. This means that the respective observation mediumof the time period and pangolin definition is always decisive for analysingSARS-CoV-2 data.