R-BIND: An Interactive Database for Exploring and Developing RNA-Targeted Chemical Probes

Abstract

While the opportunities available for targeting RNA with small molecules have been widely appreciated, the challenges associated with achieving specific RNA recognition in biological systems have hindered progress and prevented many researchers from entering the field. To facilitate the discovery of RNA-targeted chemical probes and their subsequent applications, we curated the RNA-targeted BIoactive ligaNd Database (R-BIND). This collection contains an array of information on reported chemical probes that target non-rRNA and have biological activity, and analysis has led to the discovery of RNA-privileged properties. Herein, we developed an online platform to make this information freely available to the community, offering search options, a suite of tools for probe development, and an updated R-BIND data set with detailed experimental information for each probe. We repeated the previous cheminformatics analysis on the updated R-BIND list and found that the distinguishing physicochemical, structural, and spatial properties remained unchanged, despite an almost 50% increase in the database size. Further, we developed several user-friendly tools, including queries based on cheminformatic parameters, experimental details, functional groups, and substructures. In addition, a nearest neighbor algorithm can assess the similarity of user-uploaded molecules to R-BIND ligands. These tools and resources can be used to design small molecule libraries, optimize lead ligands, or select targets, probes, assays, and control experiments. Chemical probes are critical to the study and discovery of novel functions for RNA, and we expect this resource to greatly assist researchers in exploring and developing successful RNA-targeted probes.

Graphical Abstract

Broad appreciation for the regulatory functions of RNA molecules,^1,2 along with their involvement in several human diseases,^3-5 has dramatically increased interest in developing RNA-targeted chemical probes.^6-12 At the same time, the limited chemical diversity of RNA hinders the ability of researchers to develop high affinity ligands that are also specific in a biological environment, particularly in the presence of rRNA and tRNA, which are abundant and highly structured.^1,13 Further limitations include a paucity of high-resolution structures,^14-16 limited understanding of in cellulo structures,^17,18 and few methods to assess specific target engagement.⁶ And while oligonucleotides have shown promise in select applications, RNA sequences that are not accessible or are in highly structured regions remain difficult to target.^19,20 As part of our laboratory’s efforts to address these limitations, we collected and analyzed reports of RNA-targeted chemical probes to determine the physicochemical, structural, and spatial properties of the ligands,²¹ as well as the discovery and analysis of methods employed in successful RNA-targeting campaigns.⁶ This collection, termed the RNA-targeted Bioactive ligaNd Database (R-BIND), is distinct from previously reported collections, such as Small Molecule Modulators of RNA (SMMRNA)²² and Nucleic Acid Ligand Database (NALDB),²³ in that it incorporates biological activity as a criterion and excludes rRNA-targeted small molecules. These criteria increased the level of specificity required for incorporation into the database and focused on the targeting of less abundant transcripts, ultimately allowing for the first quantification of distinguishing properties of RNA-targeted chemical probes. R-BIND is also distinct from Inforna, which matches small molecules to individual RNA motifs.²⁴ By focusing on bioactive RNA chemical probes for the first time, R-BIND complements the open access Chemical Probes Portal, which currently only contains ligands for protein targets.²⁵

Several novel guiding principles for targeting RNA in a biological system were revealed from the original R-BIND compilations, including RNA-privileged small molecule properties²¹ and effective probe discovery and characterization techniques.⁶ We anticipate that other researchers with access to these data will benefit from having the current probe information in a centralized location in order to identify additional trends and insights to further progress the field. To make this information readily accessible, we have established an online platform that presents a user-friendly interface and a unique set of tools to search and analyze the available collection (https://rbind.chem.duke.edu; Figure 1). Users can conduct a variety of searches, including (1) parameter-based searches to easily identify and evaluate R-BIND ligands based on their physicochemical, structural, and spatial properties; (2) advanced searches for a target, in vitro or biological assays, and model systems for animal studies; and (3) entire database searches for authors, abstracts, assays, conditions, experimental observations, etc. In addition, substructure- and functional group-based searches can identify structurally similar ligands within the database. Finally, a nearest neighbor algorithm has been incorporated to allow researchers to assess similarity in chemical space of either ligands within R-BIND or user-uploaded ligands. This information can be used for rational small molecule library design, hit-to-lead optimization, or as a resource for targets, probes, assays, and control experiments. In addition, users can suggest contributions to R-BIND through the website, and the database will be updated biennially by the lab. Herein, we describe updates to the R-BIND library, the development of our web-based tools, and the potential uses and applications of this platform for the field of small molecule:RNA probe discovery.

Figure 1.

Overview and applications of the RNA-targeted Bioactive ligaNd Database (R-BIND), the first compilation of RNA-targeted ligands with activity in cell culture and/or animal models.

RESULTS AND DISCUSSION

Curation and Analysis of Current R-BIND Collection.

R-BIND contains ligands that are active in cell culture and/or animal models and target non-rRNAs.²¹ The ligands must also (i) have evidence of binding to the target in vitro, (ii) have a molecular weight <2000 Da, (iii) target RNA through only noncovalent interactions, and (iv) be highlighted by the author in the conclusion. Additional details for the inclusion criteria are included in SI section 1. Aminoglycosides, covalent ligands, peptides, and oligonucleotides were excluded for several reasons, including their distinct physicochemical properties. R-BIND ligands are divided into traditional small molecules (SM), approximately 500 Da or less, and multivalent ligands (MV) with multiple RNA-binding cores connected by linker regions. The initial R-BIND collection (v1.0), containing 67 SM and 37 MV ligands, was curated to compare bioactive RNA-targeted ligands with bioactive protein-targeted ligands, using FDA-approved drugs as a proxy.²⁶ Significant differences in physicochemical, structural, and spatial parameters were identified between the R-BIND (SM) and the FDA ligands, giving insight into privileged chemical properties of bioactive RNA ligands. This collection was later updated to include reports through May 2017 (v1.1, 12 additional ligands, 8 SM and 4 MV) and was used to analyze successful design, screening, and chemical probe characterization methods, revealing a variety of traditional as well as RNA-specific strategies utilized in RNA probe discovery.⁶ The current version (v1.2) includes RNA probes through June 1, 2018, which added 23 ligands (22 SM and 1 MV). Each update requires extensive and systematic literature searching. Details of procedures can be found in SI section 1 and in the “About” section of the website.

The most recent update revealed several notable successes in RNA targeting, including an RNA-binding splicing inhibitor in clinical trials,²⁷ as well as the first long noncoding RNA (lncRNA)²⁸ and small nuclear ribonucleoprotein (snRNP)-targeted chemical probes.²⁹ The majority of these ligands were for human RNA targets (n = 18), with a single exception of a small molecule identified for a bacterial target in the past year.³⁰ Further, only a single ligand was multivalent, a relatively small pyrido[2,3-d]pyrimidine (MW = 489 Da) targeted to CUG repeat RNA.³¹ Several of the R-BIND ligands bound to frequently targeted RNA motifs (e.g., RNA duplexes, stemloops, and internal loops), although there was a notable increase in the number of G-quadruplex targeted ligands (n = 7). Another important trend in the update is the increase in ligands discovered from focused screens (n = 14), as researchers continue to utilize known RNA-binding ligands or RNA-biased properties to identify chemical probes.

Of particular note are the significant contributions in the field of targeting mRNA splicing. In 2014, PTC Therapeutics discovered an SMN-C class of ligands from an in cellulo screen. The ligands promoted the inclusion of exon 7 in the survival of motor neuron 2 (SMN2) gene with nanomolar potency and prevented motor dysfunction as well as neuromuscular deficits in spinal muscular atrophy (SMA) mouse models.³² Follow-up studies reported working models for the specificity of SMN-C ligands, suggesting two distinct binding sites on the SMN2 pre-mRNA and partial displacement or enhancement of critical RNP complexes.^33,34 The leading candidate Risdiplam (RG7916)²⁷ is currently in four clinical trials (, , , and ) for treatment of SMA. More recently, another ligand (PK4C9) was identified that also increases exon 7 inclusion of SMN2 in Drosophila models by shifting conformations of the regulatory terminal stem-loop 2 (TSL2) element.³⁵ PK4C9 was discovered using a focused RNA-binding scaffold library and an in vitro fluorescence displacement assay with TSL2. These discoveries and the previously reported Novartis small molecule in phase II clinical trials (Branaplam, ) demonstrate the extraordinary potential of targeting mRNA splicing selectively with small molecules. With the large percentage of disease-causing mutations affecting splicing,³⁶ the potential to advance more small molecules in clinical trials is substantial.

Other noteworthy contributions include targeting G-quadruplex structures in telomeric repeat-containing RNA (TERRA) lncRNA and KRAS mRNA. CK1–14, the first chemical probe for a lncRNA, was identified as a TERRA G-quadruplex stabilizing ligand from a FRET screen of natural products and related synthetic derivatives.²⁸ The current working model suggests that stabilization leads to tighter binding of TERRA to telomeric binding protein TRF2, dissociation of TRF2 from telomeric DNA, and subsequent cell apoptosis in U2OS cancer cells. Separately, two small molecules were identified to bind G-quadruplex structures in the KRAS mRNA, a gene mutated in many cancers that yields proteins considered to be “undruggable.”³⁷ With an initial emphasis on ligands that penetrated cells, the small molecules were confirmed to target KRAS mRNA in cells at <1 μM concentration and led to significant suppression of KRAS translation and apoptosis. These successes highlight the potential targetability of putative RNA G-quadraplex structures.^9,38,39 Details of other notable successes, including targets and discovery strategies, are highlighted in SI section 3.

Cheminformatic Analysis of R-BIND v1.2.

Since the original cheminformatic analysis,²¹ 30 new R-BIND (SM) ligands have been added, representing an almost 50% increase. We therefore assessed whether the original guiding principles were still appropriate to describe RNA-targeted probes. First, we compared the 20 medicinal chemistry, structural, and molecular complexity and recognition descriptors between R-BIND (SM) v1.0 and v1.2 (SI Tables 3 and 4). The average molecular weight increased slightly from v1.0 to v1.2, and consequently, the average of most of the descriptors increased as well (SI Table 4). A Mann–Whitney U test was conducted to compare the distributions of the two libraries, and no statistically significant differences were identified. In a similar manner, the spatial properties of the new R-BIND ligands were assessed using principal moments of inertia (PMI) calculations. Consistent with v1.0, the majority of the new ligands are rod-like in shape (n = 24/30; SI Table 9), and the shape distributions and library averages were nearly identical (SI Table 7 and SI Figure 2). Collectively, the physicochemical, structural, and spatial properties of R-BIND (SM) v1.0 and v1.2 are indistinguishable, indicating that even with many of the new ligands targeting novel RNA elements, the original RNA-privileged properties are still pertinent for chemical probe discovery.

Next, we compared R-BIND (SM) v1.2 to FDA-approved drugs to assess whether the differences between RNA and protein bioactives were consistent as well. As in our previous work, we filtered the FDA library by the molecular weight range of the current R-BIND (SM) to allow for more direct comparison. Subsequently, the range for the FDA library was increased from 140–590 Da to 140–666 Da, adding 41 ligands to the FDA library (SI section 1). First, the Mann–Whitney U test was used to compare the distributions of the cheminformatic descriptors (SI Table 5). The distributions of all parameters had the same statistical significance as in v1.0 except for three descriptors: molecular weight (MW), topological polar surface area (tPSA), and van der Waals surface area (VWSA), which now have statistically significant differences (P < 0.05) in distribution between the R-BIND and FDA libraries (Figure 2A). These differences could be due to the increased FDA and R-BIND library sizes, the additional FDA ligands in the upper molecular weight range, or slight distribution changes in the R-BIND (SM) v1.2 library. Notably, the highly statistically significant differences (P < 0.001) in the structural and molecular complexity descriptors are consistent with v1.0 (Figure 2B), further emphasizing the importance of chemical architecture in RNA bioactives.

Figure 2.

Box and whisker plots for (A) parameters with statistically significant differences between FDA and R-BIND (SM) v1.2 distributions and (B) representative structural parameters with highly statistically significant differences between FDA and both the R-BIND (SM) v1.0 and v1.2 distributions. The whiskers represent the 10th–90th percentile of data, and the boxes contain the middle 50% of the data. Black lines and plus signs denote the medians and means, respectively. Statistically significant differences determined by the Mann–Whitney U test are indicated as *P < 0.05 and **P < 0.001. (C) Cumulative distance distributions from the rod vertex to members of each library.

Likewise, the spatial properties of R-BIND (SM) v1.2 and FDA were compared. Despite additional ligands and the higher molecular weight range, the averages of both libraries were nearly identical (SI Table 7). A Kolmogorov–Smirnov (KS) test was utilized to compare the shape-based distributions of the R-BIND (SM) v1.2 and the FDA libraries. Analogous to R-BIND (SM) v1.0, the differences in rod, disc, and sphere distributions were highly statistically significant (P < 0.001; Figure 2C, SI Table 8) with the R-BIND (SM) library having more rod- and less disc- and sphere-like character (SI Figure 2). These results are supported by the subtriangles where the rod triangle is more highly populated, in particular the 1a and 1d subtriangles on the rod-disc axis (SI Tables 9 and 10). This analysis reinforces the original conclusion that RNA bioactive ligands are enriched in flat, rod-like shapes.

Organization of the R-BIND Database.

The graphical user interface of R-BIND is available at https://rbind.chem.duke.edu. The user interface has the following features: Browse Page, Update Section, Single Molecule View, Parameter Search, Structure Search, Advanced Search, and Nearest Neighbor Search. The Browse Page allows researchers to quickly explore the entire database, viewing the ID, name, structure, and target of each ligand. The Update Section is a central reference for each R-BIND version, including the date, number of ligands, and reference. The remaining sections are described below. All collected data are available as two Excel sheets in the SI and as a download on the Browse page. These are named “RBIND_v1.2_A,” which contains Overview, Cheminformatic, and Design and Discovery information, and “RBIND_v1.2_B,” which contains In Vitro, Cell, and Animal experimental information. Details of the data collection procedures are in SI section 2.

Single Molecule View.

The Single Molecule View feature displays general information for each ligand on the overview page and includes five tabs with additional details: Descriptors, Design & Discovery, In Vitro Assays, Cell-Based Assays, and Animal Models (Figure 3). The overview page contains the ligand name, DOIs of references, and the CAS number, if applicable. It also describes the target, the system (e.g., bacteria or virus), RNA region, RNA secondary structure, and PDB code of the RNA:ligand structure if applicable. The Descriptors tab includes the 20 cheminformatics parameters, as well as PMI shapes used to distinguish RNA and protein bioactives.²¹ The Design & Discovery tab comprises information on the screening strategy used to discover the ligand, including assay, library, and hit rate. Finally, the In Vitro Assays, Cell-Based Assays, and Animal Models tabs contain descriptions of their respective assays and observations made by the authors in the cited work. These data, along with additional details not shown, can be downloaded for each ligand individually by clicking the “Download Results” button on the Single Molecule View page. Collectively, Single Molecule View provides a central location for R-BIND data, allowing researchers to quickly gather a plethora of information and reference respective papers if additional details are needed.

Figure 3.

Example image of the Single Molecule View for R-BIND (SM) 0061. Clicking on the Descriptors, Design & Discovery, In Vitro Assays, Cell-Based Assays, and Animal Models tabs will provide additional information.

Search Tools.

Parameter Search.

The Parameter Search (Figure 4A) filters the database based on 20 cheminformatic descriptors, as well as PMI shapes. Prior to entering the search criteria, the user selects whether to conduct the search on the R-BIND (SM), R-BIND (MV), or both libraries (Figure 4B). The user can specify a maximum and/or minimum for single or multiple parameters. This allows for simple searches, such as setting a minimum molecular weight cutoff, or more complex searches such as the combination of at least three nitrogen atoms, no more than two oxygen atoms, and a molecular weight from 300–500 Da. Next to each descriptor is a pop-up button detailing the median, mean, maximum, and minimum for the parameter in the SM and MV libraries. Further, the PMI search pop-up includes a triangle with each of the shape subgroups. After the search, the selected cheminformatic and PMI descriptors are displayed on a user-friendly interface and are grouped based on the type of descriptor (Figure 4C). The columns can be sorted to further facilitate analysis. The Parameter Search allows the users to quickly filter the R-BIND libraries based on quantitative parameters, evaluate the ligands for additional trends, and perform further analyses. Additionally, the results of these queries can be exported to a .csv file containing all of the cheminformatic information in the database for the chosen ligands.

Figure 4.

(A) Options on the Search page. (B) Example image of the Parameter Search page. (C) Representative image of part of the Parameter Search results page. The page was generated by conducting a Parameter Search as shown in B. ID, Name, Structure, and Target are always displayed, and only the descriptors searched are present in the table. An average for the selected descriptors is calculated and displayed on the top row. A similar table view is displayed from the Advanced Search page.

Structure Search.

The user can perform one of two structure searches: substructure matching or functional group filtering. In the substructure match, a structure can be drawn using standard drawing tools such as atoms, bonds, and rings. Structures can also be uploaded using common structural format files (e.g., .cml, .mol, or .sdf). In either case, additional plugins are not required to draw structures. Once a structure is created or uploaded, the search retrieves all R-BIND ligands that contain the exact substructure, irrespective of stereochemistry and charge. In addition, researchers can filter based on functional groups using a predefined list of SMARTS codes, a computational language based on molecular patterns that allows unique substructures to be described.⁴⁰ The codes are classified into three different groups for accessibility: nitrogen-, oxygen-, and sulfur-containing. Further, a user can select multiple functional groups in a query, providing a more specific structural analysis. For example, all R-BIND ligands that contain either a guanidine or a guanidine and an aromatic ester can be quickly retrieved. Both of these structure-based search tools allow researchers to quickly analyze substructures of R-BIND ligands, which is particularly powerful when comparing lead small molecules or proposed derivatives.

Advanced Search.

The Advanced Search allows researchers to search all quantitative and qualitative parameters as well as perform structure searches simultaneously. Although the individual queries are ideal for searching a few criteria with a scientific question in mind, the Advanced Search tool provides researchers great flexibility on the search interface. For example, there are several additional qualitative descriptors based on text-based categorical data: Target, In Vitro Assay, Cell Line or Assay, and Animal Organism or Assay. Each of these descriptors has a text box, where researchers can enter text and retrieve all chemical probes that contain the respective inputted keyword(s). For each descriptor, a record is maintained of all possible keywords to aid in searching, which is accessed via a pop-up button. Keyword options for selected assays can further narrow searches. For example, a user can search for ligands that were successful in decreasing the “rough eye” phenotype in Drosophila models and obtain three ligands without conducting an in-depth literature search. In addition to the descriptor fields, an unlimited keyword search is included as “All Fields” that searches the entire database and can return information based on any text (e.g., journal, author, abstract, ligand name, controls, assay conditions such as buffer type). All Advanced Search results are displayed on the interface in the same table view as the Parameter Search (Figure 4C). Collectively, this feature provides a comprehensive search of RNA chemical probes and their structure, discovery, and activity on a single page with remarkable potential for discovery and exploration.

Nearest Neighbor Search.

This similarity search allows users to compare their small molecule(s) against the R-BIND (SM) database using a k-nearest neighbor algorithm.^41,42 In this method, the 20 cheminformatic parameters used to distinguish bioactive RNA ligands from FDA-approved drugs²¹ were first normalized and then plotted in 20-dimensional chemical space. The Euclidean distance⁴² of each R-BIND (SM) to all other R-BIND (SM) ligands was calculated, and the smallest nearest neighbor distance for each R-BIND (SM) was averaged. This average distance (d = 2.0505) is a conservative metric used to define “R-BIND-like” chemical space (Figure 5A). Similar to the Structure Search, users can either draw a molecule using the plugin or upload an accepted molecular file. The input small molecule is first tautomer and protomer checked at pH = 7.4, and all 20 cheminformatic parameters are calculated. Then, the properties are normalized to the average and standard deviation of the R-BIND (SM) library, and the ligand is plotted in the same 20-dimensional chemical space as R-BIND (SM). The Euclidean distances between the input molecule and all R-BIND (SM) ligands are calculated. If the distance is below the minimum average Euclidean distance (d = 2.0505) of R-BIND (SM), the input ligand is considered a nearest neighbor to that particular R-BIND (SM) ligand and is located in bioactive RNA-targeted chemical space (Figure 5B). The search results list the top 20 R-BIND nearest neighbors for the input molecule with relative distances, and the distances that fall within the defined R-BIND chemical space are bolded. For example, the input of carbazole yields hits including harmane and proflavine, the latter of which is further in chemical space from the search molecule and thus has a larger calculated Euclidian distance (Figure 5B). The top 20 list can be downloaded from the Results page, which includes the cheminformatic descriptors of the R-BIND ligands. In addition to R-BIND, a similar list of the top 20 FDA-approved drug nearest neighbors can be downloaded from the Results page. It is important to note that as more molecules are incorporated into R-BIND, the distance defining bioactive RNA chemical space will be recalculated, further refining the concept of bioactive RNA chemical space.

Figure 5.

(A) Schematic of the nearest neighbor algorithm. The R-BIND (SM) library is plotted in 20-dimensional chemical space defined by cheminformatic parameters. The Euclidean distance of each R-BIND (SM) to all of its neighbors is calculated. The smallest nearest neighbor distance for each molecule is averaged, and this distance is used to identify nearest neighbors. (B) Example of a Nearest Neighbor Search on the R-BIND website. This search provides the top 20 closest R-BIND nearest neighbors to the input molecule. Bolded distances indicate R-BIND ligands below the minimum average nearest neighbor distance. Additional significant figures and other ligand information are available via download.

With the well-accepted idea that structurally similar compounds are more likely to exhibit similar biological activity,^43,44 the results of the Nearest Neighbor Search can provide valuable information to the community and several actionable directions for RNA researchers. For example, when a chemically similar lead targets a different RNA, that RNA target can be used as an additional selectivity comparison for the researcher’s chemical probe. On the other hand, a chemically similar lead molecule itself can be used for in vitro or in cellulo comparison to the researcher’s chemical probe, serving as a positive or negative control depending on the target. Further, it is accepted that only a small fraction of all chemical space is biologically relevant and that each protein class encompasses a unique region of bioactive chemical space.^45-48 We envision a similar anecdote for RNA targets, and therefore, research stemming from or systemically exploring similarity in RNA bioactive chemical space can lead to rational, efficient approaches for chemical probe discovery. In addition to target-specific regions, we anticipate that certain regions of RNA bioactive chemical space have an increased likelihood of targeting multiple RNAs nonspecifically.⁴⁹ Future efforts in the field to explore these ideas will lead to a dramatic increase in understanding the molecular recognition principles of and specificity in bioactive RNA-targeting.

Community Involvement.

As the purpose of the database is to be a practical tool, we welcome involvement from the community. There is a Contribute section on the front page of the website for researchers to suggest new studies and ligands for R-BIND and to leave comments about the accuracy of currently available data. Changes to the database will be recorded in the Updates section of the website. This is especially useful given the time and cost involved in updating R-BIND. To assist with suggestions, we have included details on how ligands are chosen in SI section 1. We also welcome recommendations for improvements to the website and ideas for additional tools.

SUMMARY AND OUTLOOK

This work combines an expansion of the R-BIND database; an updated analysis of the physicochemical, structural, and spatial properties of R-BIND ligands; and a description of an online platform that renders the information accessible to the community and provides a number of new tools. In the expansion, there was a notable lack of new multivalent ligands, which may signal decreased interest in this targeting strategy or the need for novel approaches to multivalent ligand design. Further, most of the new chemical probes target human RNA elements, including many first in-class examples (e.g., lncRNA²⁸ or snRNP²⁹) and a nonribosomal ligand in clinical trials.²⁷ The marked decrease in new probes for bacterial, fungal, and viral elements suggests that novel targets or strategies are needed, which is particularly important for many unmet diseases.⁶ We also observed that more ligands were discovered through focused screens as the field continues to utilize RNA-privileged scaffolds and properties for probe discovery. Notably, analysis of the updated R-BIND (SM) library, which increased from 67 to 97 ligands, showed very little change in the averages or distributions of the 20 cheminformatic descriptors and PMI shapes. Comparison of R-BIND v1.2 to FDA-approved drugs revealed statistically significant differences in the distributions of some medicinal chemistry properties, and the distributions of structural and spatial properties remain highly statistically significant. Another recent report found several similar trends for RNA-binding ligands identified from a large screening library.⁵⁰ Collectively, these results support the existence of a privileged space for RNA-targeted chemical probes, which can be leveraged in RNA-targeted library design and lead optimization.

Given the features described above, we envision that the compilation of R-BIND and the tools in this platform will have several valuable applications for researchers studying RNA. To begin, researchers interested in modulating a specific biological pathway can readily identify biologically active ligands for a specific RNA. Searching or browsing the database based on target, cell line, or animal model can also yield valuable controls. For example, a researcher aiming to design a study utilizing a small molecule microarray for screening against viral RNA targets could use the Advanced Search features to identify molecules for screening (Figure 6).^51,52 Last, medicinal chemists interested in optimizing a given lead molecule could test the R-BIND-likeness of proposed derivatives while pursuing structure–activity relationships and/or identifying potential off-target effects based on the targets of nearest neighbors or similar substructures. The details of assays can facilitate the evaluation of a new lead probe by suggesting conditions and methods used for specific targets or ligand families and allowing for direct comparison of new experimental values to literature values. Further, the database can be utilized to search for successful screening assays and libraries based on particular targets or classes of molecules. It can also be used as inspiration to develop novel screening assays based on the biophysical information for a particular probe. Last, users can generate RNA-focused libraries based on scaffolds, substructures, privileged properties, and/or regions of RNA bioactive chemical space. The potential applications and subsequent impact of this platform are exceptional, and the website is a unique resource for rapidly expanding both the number and quality of RNA-targeted chemical probes.

Figure 6.

Example of an Advanced Search. A search was conducted to identify R-BIND ligands for use as positive controls on a microarray screen against viral RNA targets. Upon specifying the library, target, and functional group parameters, the user is provided with ligands that fit the search criteria. The functional groups listed are chosen for microarray immobilization. Acronyms: HIV-1 TAR, Human Immunodeficiency Virus-1 Trans-Activation Response element; JEV, Japanese Encephalitis Virus; FMDV, Foot-and-Mouth Disease Virus; IRES, Internal Ribosome Entry Site.

Moving forward, we plan to update the database biennially with our own internal searches (SI section 1) and with community suggestions. Furthermore, we hope the online format will serve as a repository and research tool for RNA-targeted chemical probe development as well as a means to evaluate the current status and needs of the field. We envision that the content and tools herein will both improve the quality and specificity of RNA chemical probes as well as spur the continued discovery and exploration of RNA-targeted chemical space. As the RNA revolution continues, chemical probes are essential tools to investigate novel targets, biology, and therapeutics, and R-BIND is a strategic resource to expedite those exciting discoveries of the future.