The Rfam database [1] is a collection of families of non-coding RNA and other structured RNA elements. Each family is defined by a multiple sequence alignment of a representative set of family members (seed alignment), a consensus secondary structure, and a covariance model [2], which integrates both multiple sequence alignment information and secondary structure information, and are analogous to the hidden Markov models used in Pfam [3] to determine the consensus sequence of protein families.

The database can be used to identify non-coding RNA elements within a nucleotide sequence of interest by searching it against the Rfam library of covariance models with the Infernal software [4]. Additionally, the database can be used to browse existing RNA families via keyword-based searches or direct access with the accession number or ID of specific families. Different pieces of information can be retrieved for each RNA family, including a descriptive summary, secondary structure information and consensus sequence, amongst many others. All of these functionalities can be accessed through the Rfam web-based interface. However, even though a RESTful API for the database is available, no client-side interface has been implemented so far to allow automated access to it, which makes any medium- or large-scale analysis a laborious and time-consuming process.

Here, we present a client-side interface to the Rfam database, enabling its programmatic access and therefore expanding the scope of the genomic analysis that can be carried out with the information provided by the Rfam database. The language of choice was R. This choice is based on the large number of tools already available in the Bioconductor project for the analysis of high-throughput genomic data. The software presented here complements these tools and facilitates the integration of the data retrieved via rfaRm within existing genomic workflows [5].

In order to demonstrate the functionalities of our package, we used rfaRm to analyze the reference genome of SARS-CoV-2 (RefSeq accession number NC_045512.2). For this, we first identified the non-coding RNA elements present in the SARS-CoV-2 genome. We then extracted and plotted information concerning the RNA families found to be present in the genome and showed how such information can be further processed with other Bioconductor packages.

Even though the genome is 29,903 bases long, such a sequence can be directly processed by rfaRm thanks to the internal splitting into fragments of 10,000 bases. We chose an overlap between consecutive fragments of 3,000 nucleotides to minimize the risk of missing hits present at the boundaries between fragments. We performed the search with and without clan competition to compare the results. An illustration of these queries implemented using R is provided hereafter:

library(rfaRm)

## Read genome from FASTA file

library(seqinr)

sars_cov_2_genome <- unlist(read.fasta("sars_cov_2_genoma.fasta", seqtype = "DNA",

as.string = TRUE))

## Convert DNA string to RNA string

sars_cov_2_RNA_genome <- gsub("t", "u", sars_cov_2_genome, ignore.case = TRUE)

## Search for Rfam families hits in the whole genome without clan competition

sars_cov_2_Rfam_hits <- rfamSequenceSearch(sars_cov_2_RNA_genome, fragmentsOverlap = 3000,

clanCompetitionFilter = FALSE)

length(sars_cov_2_Rfam_hits)

## Search for Rfam families hits in the whole genome with clan competition

sars_cov_2_Rfam_hits_2 <- rfamSequenceSearch(sars_cov_2_RNA_genome, fragmentsOverlap = 3000, clanCompetitionFilter = TRUE)

length(sars_cov_2_Rfam_hits_2)

## Search for Rfam families hits in the whole genome without clan competition

sars_cov_2_Rfam_hits_2 <- rfamSequenceSearch(sars_cov_2_RNA_genome, fragmentsOverlap = 3000, clanCompetitionFilter = FALSE)

length(sars_cov_2_Rfam_hits_2)

The search without clan competition returned a total of 7 RNA families in the SARS-CoV-2 genome (Table 1).

After extracting the IDs of all families, we performed queries for all of them to obtain a brief description with relevant information about the type and role of each RNA family:

## Extract Rfam IDs

rfamIDs_noClanCompetition <- as.character(lapply(sars_cov_2_Rfam_hits, "[[", 2))

rfamIDs_ClanCompetition <- as.character(lapply(sars_cov_2_Rfam_hits_2, "[[", 2))

## Iterate over set of Rfam IDs and retrieve a summary for each one

summary_list <- list()

for (id in rfamIDs_noClanCompetition) {

summary <- list(rfamFamilySummary(id))

names(summary) <- id

summary_list <- c(summary_list, summary)

}

All identified families were non-coding RNA elements typically found in the genome of beta-coronaviruses:

On the other hand, the search with clan competition enabled only returned 3 hits, corresponding to Sarbecovirus-5UTR, Corona_FSE and Sarbecovirus-3UTR. bCoV-5UTR and bCoV-3UTR are discarded because they are essentially the same hit as the Sarbecovirus-5UTR and Sarbecovirus-3UTR respectively, with the only difference that the latter are versions found specifically in SARS beta-coronaviruses. The Corona_pk3 and s2m motifs were also omitted, since they are comprised within the larger Sarbecovirus-3UTR.

Knowing the IDs of the RNA families of interest, more detailed information can be easily retrieved through a set of query functions. As an example, we acquired detailed information about the Sarbecovirus-5UTR (with Rfam accession number RF03120).

First, we extracted the consensus sequence and secondary structure for the family:

## Retrieve consensus sequence and secondary structure and save them to a file in the

## extended dot-bracket format

rfamConsensusSecondaryStructure("RF03120", filename = "RF03120_cons.txt", format = "DB")

## [1] "AuauuAgGcuuuuACCuaccCaGGaa..aagCcAAccAA.uuUcGauCuCUUGUaGauCUGuuCUcUAAAcGa.aCUUUAAAA……UCuGcGuggCuGUCgCucgGCUGcAUGCcuaGcGCacccaCgCaGUAUAAauAaUAAuaAAuUUUAcUGuCGuuGaCagGgaaCgaGUAACuCGuCcauCuuCuGCAGgCuGCUcaCGGUUUCGUCCGugUUGCaGcCGAUCAUCaGCacacCcAGGUUUcGUCCgGguguGaCCGAAAGGuaaGaUgGaGaGCCucGucCcuGGuuuCaaCGaGAAAA"

## [2] "……<<<<<<<.<<<. . . .>>>>>..>>>>>. . . .. . .. . ..<<<<<. . . ..>>>>>.<<<<. . .. . ..>>.>>. . . .. . . .. . .. . .<<<<<<<<.<<.<<<<.<<<. . . ..>>>.>>>>>>.>>>>>>>>. . . .. . . .. . . .. . . .. . . .. . . .((((((((((((.(((((. . .(((.(((.((((<<<..<<<<<<.<<<<<. . .. . .>>>>>..>>>>>>. . .. . .>>><<<<<<<.<<. . .. . .>>>>>>>>><<<. . . .>>>)))).)))))).))))))))))…)))))))… .."

Next, we extracted the seed multiple sequence alignment of the family:

## Retrieve seed multiple sequence alignment and save it to a FASTA file

rfamSeedAlignment("RF03120", filename = "RF03120_seedAlgn.fasta", format = "fasta")

The information from the consensus secondary structure and the multiple seed alignment can be combined and visualized with the R4RNA package (Fig 2):

## Read consensus secondary structure and multiple alignment

library(R4RNA)

library(Biostrings)

secondaryStructureTable <- readVienna("RF03120_cons.txt")

seedAlignment <- readBStringSet("RF03120_seedAlgn.fasta")

## Make a helix plot of the consensus secondary structure and annotate it with information

## from the seed alignment

plotCovariance(seedAlignment, secondaryStructureTable, grid = TRUE, line = TRUE, arrow = TRUE,

legend = FALSE, cex = 3)

It is also possible to generate plots of the secondary structure annotated with different types of information by specifying the “format” argument. Possible values are:

For example, in order to generate a plot of the secondary structure of the Sarbecovirus-5UTR annotated with sequence conservation (Fig 1B), the following example code can be used:

## Generate a plot of the secondary structure of Rfam family RF03120 annotated with

## sequence conservation and save it to a PNG file

rfamSecondaryStructurePlot("RF03120", filename = "RF03120_cons.png", plotType = "cons")

## The plot can also be saved to an SVG file, which is useful to save images into an

## editable vector-based format

rfamSecondaryStructureXMLSVG("RF03120", filename = "RF03120_cons.svg", plotType = "cons")

A plot of the phylogenetic tree of the seed alignment can be easily generated and labeled with species names or sequence accession numbers (Fig 1C):

## Generate a plot of the phylogenetic tree of Rfam family RF03120 labeled with species

## names and save it to a GIF file

rfamSeedTreeImage("RF03120", filename = "RF03120_tree.gif", label = "species")

Additionally, the phylogenetic tree can be retrieved in the New Hampshire Extended (NHX) format. The tree can then be read and processed with other software, such as the treeio R package:

## Save the phylogenetic tree of Rfam family RF03120 to a file in the NHX format

rfamSeedTree("RF03120", filename = "RF03120_treeNHX.nhx")

## Read the tree as a treedata object

library(treeio)

treeioTree <- read.nhx("RF03120_treeNHX.nhx")

## Print a summary of the tree

as.phylo(treeioTree)

##

## Phylogenetic tree with 19 tips and 17 internal nodes.

##

## Tip labels:

## _DQ022305.2/1-295_Bat_SARS_coronavirus_HK…1, _DQ648857.1/1-297_Bat_CoV_279/2005.1, _MG772934.1/1-298_Bat_SARS-like_coronavirus_{}.1, _MT345841.1/1-293_Severe_acute_respiratory_syndrome_coronavirus_2.6, _MT344963.1/1-299_Severe_acute_respiratory_syndrome_coronavirus_2.2, _MT345869.1/1-293_Severe_acute_respiratory_syndrome_coronavirus_2.5,…

## Node labels:

##, 0.780, 0.940, 0.800, 0.890, 1.000,…

##

## Unrooted; includes branch lengths.

The rfaRm R package provides an easy-to-use client-side interface to the Rfam database, enabling users to access it programmatically and therefore bypassing the limitations of interactive access through the web interface or the requirement to install and search the database locally. Programmatic access allows the identification of non-coding RNA across entire genomes. The package is designed to interoperate with existing software by returning data into formats directly readable by other tools and R packages and is available as part of the Bioconductor project, which facilitates its integration within workflows and pipelines for the analysis of genomic data. We believe the package will provide a useful resource for the community of RNA Bioinformatics, whose interest in the tool has been demonstrated by the considerable number of downloads of the package in spite of its still short lifetime.