The most widely accepted DNA structure is the classical B-DNA form, which is a right-handed double helix in nature and contains the hydrogen bonds between the nucleobases as described by Watson and Crick in 1953 (Watson and Crick, 1953). Yet, it is evident that DNA is structurally dynamic and can also adopt alternative secondary structures like guanine-rich tetrads (Gellert et al., 1962) and non-guanine rich tetrads (Liu et al., 2018). Guanine-rich DNA strands are capable of folding into four-stranded helical structures, called G-quadruplexes (G4s). The four guanine residues present in the core of a G4 are bonded through the Hoogsteen hydrogen bonds and are stabilized by monovalent cations like K⁺ and Na⁺ to attain a planar form (Gellert et al., 1962; Sen and Gilbert, 1990). A minimum of three consecutive guanine residues is stacked four times successively (called G-tracts or G-runs) with the intervening sequences extruded as loops and taking the shape of a G4 scaffold (Bochman et al., 2012). It is believed that longer perfect G-tracts increase the stability of the G4 structure while unusual G-tracts tend to decrease the stability of the structure (Huppert and Balasubramanian, 2005). On the contrary, new models have predicted the robustness of the unusual G-tracts (Varizhuk et al., 2017; Doluca, 2019). Furthermore, in the early models of G4 structures, it was assumed that loop lengths up to 7 bases could form stable quadruplexes (Huppert and Balasubramanian, 2005). However, it has been since then observed that the quadruplexes with several long loop lengths (up to 30 bp) could also exist (Guedin et al., 2010). It has been shown that RNA can also adopt this type of non-canonical structure under physiological conditions (Davis, 2004). G4s can be unimolecular or multi-molecular and can attain a variety of topologies arising from different combinations of strand direction, length, and loop composition and both intramolecular and intermolecular G4s can also be observed (Spiegel et al., 2020). During the last few years, there has been a lot of growing interest in the scientific community in exploring these G4 structures and their regulatory roles among biological processes (Maizels, 2006; Rhodes and Lipps, 2015; Hansel-Hertsch et al., 2017; Spiegel et al., 2020). They are widely distributed in prokaryotes, eukaryotes, and viruses and play key roles in regulating several physiological and pathological processes. Some of the biological processes known to be regulated by G4 structures include DNA replication (Valton and Prioleau, 2016), damage and repair system (Fleming et al., 2017), genomic instability (Wang et al., 2019), gene expression (Cave and Willis, 2022), chromatin rearrangement (Reina and Cavalieri, 2020), and viral latency (Ruggiero and Richter, 2018). Their role in diverse biological processes renders them interesting potential therapeutic targets (Ruggiero and Richter, 2018; Carvalho et al., 2020). Different studies in various organisms have revealed that G4 secondary structures are located in a non-random manner within genomes and tend to cluster in particular/functional genomic regions like telomeres (Sundquist and Klug, 1989), promoters (Siddiqui-Jain et al., 2002; Dai et al., 2006; Fernando et al., 2006; Xu and Sugiyama, 2006), and untranslated regions (UTRs) of mRNA (Huppert et al., 2008). The role of G4s is associated with several diseases including cancer (Varshney et al., 2020), neurogenerative disorders (Wang et al., 2021; Vijay Kumar et al., 2023) and rare genetic disorders including fragile X syndrome (Asamitsu et al., 2021).

Initial in silico approaches for the prediction of putative G4 structures on a genome scale were based on the experiments conducted biophysically (Huppert and Balasubramanian, 2005; Todd et al., 2005). The ongoing era of next-generation sequencing has made whole genome sequencing relatively easy and affordable thereby creating a wealth of genomic data which can be used to obtain a bird’s eye view of the cellular processes. Using techniques like rG4-seq and G4-seq, scientists have developed the transcriptome-wide (Kwok et al., 2016), genome-wide (Chambers et al., 2015) experimental map of G4s in humans, and recently, the more exhaustive whole-genome landscape of G4s in 12 species (Marsico et al., 2019). Several improved computational algorithms have been developed by employing the G4-seq dataset to train a machine learning model for the characterization of G4 structures on a genomic level (Garant et al., 2017; Hon et al., 2017; Sahakyan et al., 2017). Notably, the vast majority (80%–90%) of the G4 structures predicted by these improved computational approaches were confirmed to exist in genomes by the G4-seq approach (Marsico et al., 2019). A powerful tool called pqsfinder (Hon et al., 2017) provides a flexible framework for its users and allows them to define the custom criteria for scoring and matching. It allows the user to input up to three imperfections (mismatches, bulges in G-runs, and/or long loops >9 nt) in a single sequence of DNA or RNA and has the advantage of assigning a score to each predicted G4 sequence. The scoring scheme emphasizes the stability of the predicted structure because it gives a bonus score to the perfect G-tetrad stacking and a penalty score in case of mismatch and bulges. Quadron (Sahakyan et al., 2017) is a machine learning (ML) model based on a tree gradient boosting machine and trained on the G4-seq data for the human genome, which allows the user to predict G4 structures in DNA as well as RNA sequences. G4RNA screener (Garant et al., 2017) applies an ML model based on an artificial neural network and trained on experimentally validated G4s from sequences deposited in the G4RNA database. It allows the user to predict G4s in RNA sequences only and incorporates the cG/cC and G4 hunter algorithms for better or comparable outcomes. Additionally, ImGQfinder (Varizhuk et al., 2014) is another tool where a user can predict G-quadruplexes. This tool allows one mismatch or bulge in G-tract. QPARSE (Berselli et al., 2020) is a graph-based search algorithm where users can look for monomeric and multimeric quadruplex forming sequences and G4s with long, hairpin loops. Users are allowed to enter the query sequence of a maximum 10,000 bp length or upload a fasta file of a maximum 15 Kb size. G4-iM Grinder (Belmonte-Reche and Morales, 2020) looks for G4s and i-Motifs within a given DNA or RNA sequence. It has three distinct methods: the G4 search engine with 13 customizable functions (for example, showing G4 on both strands, loop sequence, size, etc.), G4 qualification functions, and quantification functions. It incorporates cG/cC and G4 hunter algorithms to evaluate better results. The continuous progress in literature providing evidence on the in-vitro existence of G4 structures containing more than four G-tracts (Phan et al., 2005; Omaga et al., 2018) and G4 structures containing all the possible tetrads, A:T:A:T tetrads and bulged nucleotides in one single structure (Liu et al., 2018) still remained to be incorporated into the search algorithm.

Anotpther interesting role of G4 structures is to influence the methylation at CpG islands (CGIs), which are guanine-cytosine-rich regions and are usually hypomethylated. The CGIs are widespread at the promoters of housekeeping, tissue-specific, and developmental genes and co-localize with G4s in these actively transcribed regions for gene regulation (Jara-Espejo and Line, 2020). Recently, it was proposed that the G4 structures protect the CGIs from methylation by sequestering and inhibiting DNA methyltransferases and hold an important place in epigenetic control mechanisms (Cree et al., 2016; Mao et al., 2018). As both G4s and CGIs are tightly associated with actively transcribed regions, their accurate identification in the genome is of great significance.

Several computational tools for the identification of CGI in a given DNA sequence are accessible to users nowadays. Three of the widely used conditions for CGI analysis are as follows 1) moving window should be of 200 nucleotides, 2) GC content higher than 50%, and 3) CpG O/E (Observed/Expected) ratio higher than 0.6 (Gardiner-Garden and Frommer, 1987). Some improved versions for CGI identification are also available which include the additional parameters (Larsen et al., 1992; Ponger and Mouchiroud, 2002; Takai and Jones, 2003; Hackenberg et al., 2006).

The presence of stable and conserved G4s in all known human viruses and their variants has been successfully presented by Lavezzo et al. (2018) but their analysis is mainly based on the reference genomes (RefSeq). Additionally, the analysis pipeline has failed to provide any criteria where user-defined sequences can be handled. It is evident that mutations in viruses have played a major role in evolution (Moelling and Broecker, 2019) enabling them to evade host immune responses efficiently (Xia et al., 2018). Several studies related to mutations in quadruplex-forming structures have shown that mutations in G4s may hinder normal cellular activities (Khristich and Mirkin, 2020). Therefore, understanding the potential effects of mutations on quadruplex-forming structures warrants further research.

All quadruplexes and CGIs search models have limitations despite the advancements in the field as none of the tools has been explicitly designed to detect and analyse all possible G4s. Keeping in mind that G4 forming sequences within a genome harbour CpG sites, we have developed a web-based server, QUFIND (QUadruplex FINDer), where users are allowed to predict either the G4 structures or CGIs and G4 structures simultaneously, in a wide range of organisms. This is the first-ever tool that allows users to find the G4 forming sequences within CGIs so that the user can analyse G4 secondary structures in context with the CGIs. The web server interfaces with ENSEMBL’s REST API and NCBI API to mine CGIs and/or G4 secondary structures. It provides options to search the entire ENSEMBL and NCBI databases in order to retrieve the desired nucleotide sequence entries for analysis. The web server is organized in such a way that users are free to choose as many possible parameters by themselves. The web program is divided into three modules: QUFINDU which allows querying CGIs and/or G4 secondary structures in all the species for which the sequences are available in the ENSEMBL and NCBI database, QUFINDV which allows querying G4 secondary structures in viruses and their variants and lastly, QUFIND which allowed CGIs and/or G4 secondary structures to search in user-provided sequences. QUFINDV offers an interactive graphical representation of the G4 sequences in viruses and their variants for comparative studies on a single screen. The program is also designed to handle the analysis of non-guanine-based quadruplexes as well as the prediction of overlapping G4 structures among the CpG islands.

QUFIND is the first web server that allows users to find quadruplexes either without or within the CGIs. The tool provides an opportunity for the users to analyze G4s in the context of the CGIs. In addition, the server is able to mine G4s in viruses and their variants. This web server is connected with ENSEMBL’s REST API and NCBI’s e-utilities to get the latest gene models or genomic assembly of any organism to detect the presence of secondary structures in them. This server also provides an interactive graphical representation of the mined secondary structures and the results can be downloaded in a convenient format.

The methodology of secondary structure mining allows flexible customization of stem length, loop length, nucleobase, and inclusion of defects. Generally, motifs with three or four nucleic acid-tracts and a loop length of 7 nucleotides are considered to be more stable but due to continuous developments in literature, unusual G4-forming structures can also be seen. Hence, QUFIND is meant to be a flexible and comprehensive tool for investigating G4s.