We downloaded mutation data from the NCBI’s SRA using the prefetch feature to download SRA files. Selection of files was complicated by divergent numbers of samples submitted in different projects, so we selected them from several labs across the world, which provided large batches of samples over extended periods of time. We reasoned such labs would have likely perfected the sequencing process and provided reliable samples for different VOCs.

Thus we obtained >35,000 samples and filtered them as follows. Samples under 10MB in size were excluded as they had low read depth overall, preventing reliable detection of single-nucleotide variants (SNVs). We selected samples with <700 nt of zero read depth over the whole genome, to avoid effects of biased coverage and of lack of coverage for potential SNVs ( Supplementary Figure 1 ). To process the data in consistent way, we used fastq files (when available) or unaligned the reads from BAM files using Samtools (9) to convert them to fastq files. We then aligned the fastq files to the SARS-CoV-2 reference sequence (10) (NC_045512.2) using BWA mem.

We then used GROM (11) to find SNVs in the data. In total we found ~25,000 unique SNVs using GROM across >25,000 samples. GROM was run using default settings and the “remove duplicates” option, to minimize PCR duplicates. Relevant data from output VCF files were then consolidated to SQL files detailing the sequencing platform, coordinates, and alternate nucleotides for each sample.

To identify repeats in the SARS-CoV-2 genome, we analyzed the Wuhan reference sequence (10) using UGENE (12) with the default settings and selected repeats >5 nt long. We then identified super-repeats of one another (superstrings of shorter repeat strings) using in-house scripts.

Table 1 describes datatypes, sources, and tools used to generate the data that populates the database. Protein structures were obtained from the Zhang group who has used I-TASSER to predict protein structure for all SARS-CoV-2 proteins (13). Their predictions are highly accurate for the SARS-CoV-2 proteins despite relatively few homologous sequences with available protein structures.

To show the secondary structure of SARS-CoV-2 genomic RNA we collected several datasets from groups that have measured the viral RNA accessibility at a single base resolution. The first of these was taken from Manfredonia et al. (16) who has used SHAPE and DMS mutational profiling to find secondary structure maps with single base resolution. Yang et al. (14) has used SHAPE-MaP to find the reactivities of the reference sequence as well as a delta variant sequence. Finally, Sun et al. (15) has used icSHAPE to map reactivities.

Data presented as Intragenome Interaction Data represent regions of pairwise RNA interactions across the genome. Such regions have been detected via proximity ligation sequencing was performed using SPLASH to find these regions in Vero-E6 infected cell (14).

We obtained the coordinates of viral non-coding regions, its genes and proteins, their respective nucleotide and amino acid sequences from the NCBI record of the SARS-CoV-2 (NC_045512.2). SARS-CoV’s information was retrieved in the same way from the NCBI record of the Tor reference sequence (NC_004718.3). We then performed a thorough literature review (across hundreds of papers) of proteins in SARS-CoV-2 and SARS-CoV to obtain their functional descriptions. Next, we identified the available coordinates of functional domains in both viruses. Using BLAST (17) and CLUSTAL-W (18), we further performed pairwise alignments of the proteins of SARS-CoV-2 and SARS-CoV to evaluate the levels of amino acid identity of the homologous functional domains. We manually curated mismatched coordinates of such homologous domains between different studies, produced reconciled coordinates and transferred the domain annotations, further accompanied on respective pages by the publications describing them.

Users can access SARSNTdb at https://grigoriev-lab.camden.rutgers.edu/sarsntdb/

The website is implemented in PHP (version 7.4.29) and the SQL server through mysql (version 15.1) with MariaDB (distribution 10.3.34).

The interface of the database consists of several tabs. The Search tab has a dropdown menu that brings the user to a Genome Search, Mutation Search, and a Repeat Search. These searches are interconnected to allow the user to take information gleaned from one search into another. The Help tab instructs the user on how to use the website by providing an example. The Reference tab brings the user to this article where they can learn about the data sources and how the website was constructed.

The Genome Search page allows the user to specify nucleotide coordinate intervals and find information about functionally relevant regions of the SARS-CoV-2 virus that overlap or are contained between these coordinate pairs. Such regions most often correspond to genes and functional domains they encode. Also, this search reports about nearby repeats and intragenomic interactions obtained using a SPLASH technique (14).

One can also select a single ORF or Nsp from a menu to get to such genes. Their protein products are described on the Protein Detail page ( Figure 2 ). In addition to images of the predicted structure of the SARS-CoV-2 proteins, their functional domains, smaller motifs, and certain amino acid residues with annotated functionality are also displayed graphically. At the bottom of this page there are the relevant RNA and Protein sequences derived from the respective NCBI reference.

Similar to evaluations of SARS-CoV-2 mutations using variant effect predictors (19), a link to online analysis of all possible amino acid substitutions for a given protein by SNAP2 (20) is also provided here. A single click would copy the protein FASTA sequence and redirect to the SNAP2 server, so users just need to paste it there, start the analysis and view its results.

Knowing that a substitution occurred in a protein domain with certain function may provide more specific information on the functional effect. Since SARS-CoV-2 domains are typically derived from the previous body of work on SARS-CoV, we devoted a special page for each protein in both viruses for comparing domains. This page is linked form the Protein page and contains a table detailing the similarities of the two viruses and an alignment of both protein sequences created using CLUSTALW (18) and BLAST (17). The coordinates in the table are derived from primary literature and review papers (that can be accessed by clicking the hyperlinks on the coordinates) and sometimes they differ, despite being reasonably well aligned. These pages also illustrate the degree of conservation of the two viruses and links to mutations for each domain are also provided.

The Mutation Search page allows the user to search for mutations in a nucleotide range or within a gene. The search results are bar graphs depicting the number and type of substitutions in the range. The bars are also subdivided by sequencing platforms. If a more granular view of the mutations is needed the user can click on Mutation Detail to see expanded information related to the mutations in the that nucleotide range. Below the Substitution Frequency table there is a histogram that displays SNV frequency across the nucleotide range selected.

Also on this page is SHAPE data that may help inform the user why certain regions may be conserved due to the secondary structure constraints. When the size of the searched region is large, the SHAPE Data is displayed in intervals where the SHAPE value is averaged across that region. If the size of the searched interval is under 100 nucleotides, each position and the SHAPE value is displayed individually. If the shape value is above 0.5 it is displayed in blue indicating a high reactivity while below 0.5 is displayed in red and indicates a low reactivity. We selected several SHAPE datasets and displayed them in separate graphs. These data make up our Mutation Search page and it is visualized on the page using CanvasJS (Fenopix Inc.).

The Repeat Page ( Figure 3 ) allows the user to search the SARS-CoV-2 reference sequence for repeats of size 6 nucleotides or greater. Displayed on this page is the genome schematic with proteins colored distinctly. When a repeat is found red lines appear on the genome indicating repeat locations, and a table displaying the coordinates of the repeats as well as which protein they appear in is displayed. Also available are repeats, which are super-strings containing the searched repeat; these are deemed super-repeats. For example, the repeat AACAGGA is a super-repeat of AACAGG as the former is a super-string of (i.e., contains) the latter. Clicking on these super-repeats brings the user to a Repeat Page for the super-repeat (with their respective super-repeats, if available). For the default search on the Repeat Page and a clear biological example, we provide the minimal repeat of the transcription regulatory sequence (TRS) from the SARS-CoV-2 virus (5), with all locations of canonical TRS visualized throughout the genome for the user.

As stated previously there are many databases tracking the waves of VOCs and their typical mutations. The virus continues to evolve, and even the general public is made aware of new substitutions in the best-annotated spike protein. When new mutations appear, it is important to be able to quickly identify where they occur and analyze their effects by detecting genome features nearby. Furthermore, substitutions take place not only the spike protein, yet those affecting other parts of the genome are typically ignored in the databases and many analyzes.

In contrast, SARSNTdb could be an excellent starting point for such quick evaluation, and a schematic walkthrough is shown in Figure 1 (bottom). Consider the mutation C28311T, found in Omicron. Let us first go to our Genome Search page and input the coordinate 28311 and find it is part of two overlapping genes, encoding the Nucleocapsid (N) protein as well as ORF9b. In N it is located in the N-terminal arm/Intrinsically disordered region. In ORF9b we see it is part of the site that interacts with the host protein NEMO. We also see that it is a part of some common repeats and has intragenomic interactions at the 5’ end of the protein as well as a region 200nt away that it binds with. These close intragenomic interacting regions could form pockets that may become therapeutic targets (15). Clicking View Details for ORF9b, we find its function and see it supresses the innate immune system through regulating Mitochondrial Antiviral Signaling pathways (21). In comparing it to SARS-CoV we find this domain is not well conserved with only 63% similarity overall. In the paper linked via the domain coordinates in the table, we find that this region, when deleted, resulted in a loss of function of the protein and its interaction with NEMO (21). If this nucleotide change results in a non-synonymous mutation, it could affect the function of the protein. By clicking Mutations on the table, we are brought to the mutation page showing the NEMO interaction region’s mutation frequencies, SHAPE scores and, if we click Detail, a breakdown if that specific mutation has been found. If it has been found the detail page will also show the type of variant it creates. In this case the SNP has been found previously in thousands of samples, where it changed a proline to a serine. In addition, the SHAPE score of this nucleotide is low according to all datasets, indicating that it may be prone to forming secondary structures within the RNA. Overall, with all such results about this mutation we can conclude that it should be monitored as it has been persisting over time and now, with Omicron spreading rapidly, may be gaining increased prevalence. This mutation could affect the ability of ORF9b to supress the innate immune system through interacting with NEMO and its effects should be explored further.

SARSNTdb is a database for users of varying levels of knowledge about virology or genomics. It provides nucleotide-level functional information about various aspects of the SARS-CoV-2 genome. It features a quick and easy coordinate-based search for SARS-CoV-2 gene and protein functions, mutations found in patient samples, structural and sequence elements of the virus RNA and several other features. We reviewed, analyzed, and provided visualization for data that could help users to better understand the virus, and to do this rapidly. We will continue to add mutation data as we process other representative samples from the NCBI to detect SNVs in those with GROM on a regular basis. In addition, should new domain definitions be discovered they will also be added to the database.