Silencing with precision: In silico strategies for antisense oligonucleotide engineering

Abstract

Therapeutics based on nucleic acids are emerging as a transformative drug class in the era of personalized medicine. They include antisense oligonucleotides (ASOs), aptamers, small interfering RNAs (siRNAs), small activating RNAs (saRNAs), microRNAs (miRNAs), and long non-coding RNAs (lncRNAs), in addition to messenger RNAs (mRNAs). ASOs are short synthetic molecules of nucleic acids designed to bind to specific RNA molecules to allow for precise modulation of gene expression in a personalized medicine context. This review will summarize the state of the art in in silico tools and methods that are used for the design, 3D structural prediction, molecular docking, and simulation of ASOs, including their applications and limitations. By leveraging these computational approaches, the review will provide in silico methodologies toward improved specificity and efficacy of ASO therapeutics. These methodologies can enhance understanding of biological mechanisms of ASOs to allow for improved accuracy and efficacy of ASOs in the clinic while also providing new pathways toward the development of targeted therapies.

Introduction

Antisense oligonucleotides (ASOs) are a breakthrough in molecular medicine, providing a specific method of gene expression modulation. The idea of ASOs was initially conceived by Zamecnik and Stephenson in 1978, who showed the potential of ASOs to suppress the replication of Rous sarcoma virus (RSV) by binding to its RNA [1]. More theoretically specific than traditional small molecule-based drugs, ASOs are artificial short DNA oligobases synthesized to bind specifically and with high affinity to Watson–Crick base-paired complementary DNA or RNA strands, and in doing so, inhibit the translation of disease-causing proteins [2, 3].

ASOs present major benefits over small molecules and monoclonal antibodies, such as high specificity in binding, simpler synthesis, improved penetration within cells, and reproducible manufacture [3]. With a reduced molecular weight, they are capable of specifically targeting RNA sequences without off-target activity, which provides superior therapeutic effectiveness, particularly for genetic diseases. ASOs can also be modified chemically for greater stability and effectiveness, with ASOs as a highly suitable and promising second choice in targeted therapy for other diseases [4].

The recent integration of in silico tools has profoundly enhanced ASO development by enhancing the evaluation of ASO-mRNA complex stability and optimizing chemical modification, which improves therapeutic effects and decreases experimental costs. Although these tools are well established for small-molecule drugs, they are not fully developed for nucleic acid-based drugs such as ASOs, owing to the complexity of nucleic acids [6–8].

As the use of ASOs becomes more prominent as a therapeutic modality, there is a growing necessity to streamline their design and action to improve their efficacy and reduce undesirable off-target effects. In silico approaches provide an effective remedy in this respect by allowing researchers to anticipate and screen for the interactions between ASOs and the targeted RNA sequences, thereby aiding in the creation of more efficient ASO-based treatments. This article intends to emphasize the central role of computational methods in ASO investigations, highlighting their uses, strengths, and limitations [6, 9].

Basic structure and functional mechanism

ASOs typically consist of a structure based on a backbone consisting of a chain of nucleotides joined by phosphodiester bonds. However, to enhance stability, specificity, and resistance to enzymatic degradation, ASOs are commonly chemically modified [5, 11, 12]. Common modification methods include the use of a phosphorothioate backbone that substitutes a non-bridging oxygen atom with sulfur and the addition of 2’-O-methyl or 2’-O-methoxyethyl modifications of the ribose sugar. These modifications stabilize ASOs against nuclease degradation and have a better binding affinity for target RNA, thus resulting in improved efficacy in therapeutic applications [12, 13].

The way ASOs function largely relies on their ability to regulate how genes are expressed through attachment to specific sections of RNA [10]. There are many types of ways that ASOs work. One common way is through the process called RNase H cleavage. In this method, an ASO would be connected to a specific area of an RNA molecule, forming a hybrid DNA-RNA complex that RNase H would then recognize and cut apart, causing the destruction of the RNA that it is targeting. Another method is called steric inhibition. Steric inhibition occurs when an ASO attaches to the start of an mRNA, preventing the ribosomal complex from attaching to it and translating it into a protein. ASOs can also function as splicing modulators because they can attach to pre-mRNA and change whether an exon gets included or excluded in the final processed mRNA molecule. The various modes of action of ASOs provide researchers and clinicians both opportunities for conducting research related to manipulating gene expression and being able to treat diseases through gene therapy (Fig. 1) [8, 14, 16].

Fig. 1.

The diagram illustrates the mechanism of action for ASOs. In the nucleus, ASOs hybridize to pre-mRNA and regulate splicing to yield modified mature mRNA. In the cytoplasm, ASO-mRNA hybrids are bound by RNase H1, causing cleavage and degradation of mRNA. ASOs also sterically hinder ribosome binding, inhibiting mRNA translation into proteins

Mode of action of ASOs

ASOs use several different ways to affect gene activity, and each has its own unique therapeutic applications. A major mechanism of ASO action is through the RNase H1-mediated silencing pathway. In this process, ASOs "complement" the target mRNA by forming Watson–Crick base pairs on its RNA sequence, which leads to cleavage of the target RNA by RNase H1, resulting in decreased protein synthesis [13]. An example of an alternative mechanism of ASOs’ biological actions is through their ability to modulate mRNA splicing patterns. This form of action occurs when ASOs hybridize with regulatory sequences in the pre-mRNA to alter the normal splicing events associated with that gene, allowing for the correction of abnormal patterns of exon inclusion/exclusion to restore normal splicing processes. Examples of splice modulation are already in use as a treatment for spinal muscular atrophy (SMA) and Duchenne muscular dystrophy (DMD) [14, 15]. Finally, there is the steric block mechanism, in which ASOs block RNA–protein interactions by physically blocking and not degrading RNA, and allow splicing or translation to be regulated. Overall, these mechanisms allow ASOs to be a powerful and multifaceted tool in precision medicine [17].

RNase H1-mediated silencing

RNase H1-mediated gene silencing is an important pathway through which ASOs induce gene knockdown with ASOs designed to hybridize to sequences of mRNA, resulting in DNA-RNA hybrid formation upon association. RNase H1, an enzyme, identifies these hybrids and cleaves the RNA strand, inhibiting protein translation and effectively silencing the gene [8, 13]. This technique is very effective and specific, and it has great potential in the development of antisense drugs.

This process has been extensively used to treat conditions such as familial hypercholesterolemia and cancer, where ASOs hybridize with mRNAs of toxic proteins, inhibiting their expression. The design of good ASOs is complicated by the complexity of predicting their conformation, binding affinity, and off-target effects [18]. Substitutions in molecules [such as by adding phosphorothioate backbones or 2’-OMe] can enhance the ability of ASOs to remain intact longer [i.e. to provide stability], but these additions may also change the shape of the ASOs, which can change how effective ASOs are in producing results. Despite this, ASOs have achieved tremendous potential to effectively treat genetic disorders and neuromuscular disorders through precise regulation of levels of messaging related to genes [11, 12].

Steric block mechanisms

ASO splicing modulation constitutes a major mechanism of action that allows for the selective modification of alternative splicing events from pre-mRNA. By modifying an ASO’s design to enable the binding of the ASO to specific splice acceptor sites, splice donor sites, or to the exon splicing enhancer and intron splicing silencing sequence, the potential exists to regulate which exons are included in the mature mRNA produced by splicing. As a result of this ability to modulate splicing by targeting specific sites, ASOs are highly beneficial for therapeutic uses under conditions that result in pathogenic splicing. For example, in spinal muscular atrophy (SMA), the ASO medication nusinersen targets an ISS in the SMN2 gene, enhancing the inclusion of exon 7, which leads to the production of functional SMN protein. By correcting SMN2’s splicing pattern, nusinersen offsets the loss of the SMN1 gene, successfully treating the root cause of SMA [20–22].

Splice modulation by ASOs is also potentially useful for the treatment of other genetic diseases in which splicing defects are involved. In Duchenne muscular dystrophy (DMD), ASOs like Eteplirsen are engineered to cause the skipping of certain exons during the splicing of the dystrophin pre-mRNA, which restores the reading frame and leads to the production of a partially functional dystrophin protein. This exon-skipping approach does not fully restore normal protein function but reduces the severity of the disease substantially. In addition to genetic disease, ASO-mediated modulation of splicing is also of interest for treating cancer, since splicing regulation of oncogenes or tumor suppressor genes may prevent cancer cell growth or make cancer cells more susceptible to therapy. The ability of ASO-induced splicing modulation to be applied so broadly to modulate splicing highlights its use as a potential therapeutic tool with broad applicability to many disease states [23, 24, 26].

Steric block is a mechanism of action employed by ASOs to regulate gene expression by physically blocking the binding of cellular machinery to target RNA sequences without triggering RNA degradation [16, 17]. In this mode of action, ASOs target specific regions of the target mRNA, e.g., the 5’ untranslated region (UTR), coding region, or the splice sites, where they are capable of effectively obstructing ribosomes, splicing factors, or other regulatory proteins’ access to the targeted positions. The steric block ASOs, in such areas, will disrupt mRNA translation into protein or alter the way pre-mRNA splicing occurs. This provides a means to regulate individual proteins. This approach can be very useful when the desired outcome is the alteration of gene expression without the silencing of genes by means of RNA degradation, providing an effective way to change gene expression with minimal off-target effects [2, 3, 27, 32].

The application of steric block ASOs to block aberrantly spliced RNA is one example of this type of therapy. DMD [Duchenne muscular dystrophy] has utilized steric block ASOs to target the dystrophin pre-mRNA and block the exons being used in the splicing process [26–28]. By using the exon-skipping strategy, the reading frame is restored, and a shortened dystrophin protein is generated. The truncated dystrophin can still provide partial function for improved treatment of DMD. The utility of steric block ASOs has been extended to the treatment of tumors by blocking the synthesis of oncogenes, which leads to a reduction in the amount of protein produced that drives tumor growth [31]. The use of steric block ASOs for precision medicine and their flexibility makes them one of the most valuable tools in developing targeted therapies for a broad range of inherited or acquired diseases [17, 24, 28].

Advancements in ASO technology and comparative effectiveness with RNA interference (RNAi)

Recent advances in ASO technology involved chemically modified oligonucleotides with increased stability, binding affinity for their complementary mRNA, and increased resistance to degradation from nucleases [30–32]. The modifications have included phosphorothioate backbones, 2’-O-methyl, and locked nucleic acids. In addition to improving drug pharmacokinetics, these modifications help to decrease their immunogenic potential and toxicity. The antisense therapies’ recent successes, such as Nusinersen for spinal muscular atrophy and Eteplirsen for Duchenne muscular dystrophy, demonstrate the ability of ASOs to impact precision medicine. Research into their use is ongoing, with efforts to overcome the challenges of delivery, off-target effects, and long-term efficacy [32–34, 37]. ASOs and small interfering RNAs (siRNAs) both enable sequence-specific RNA silencing but differ fundamentally in their molecular architecture and mechanisms of action. ASOs are generally single-stranded oligonucleotides that directly hybridize to a complementary RNA target and can induce RNase H-mediated RNA degradation or sterically interfere with RNA processing or translation. In contrast, siRNAs are double-stranded RNA molecules in which the guide strand is incorporated into the RNA-induced silencing complex (RISC), where Argonaute-mediated recognition and cleavage of the complementary target RNA results in gene silencing. These mechanistic differences can influence cellular localization, duration of activity, specificity, and biological responses (Fig. 2).

Fig. 2.

Comparison of ASO- and siRNA-mediated gene silencing mechanisms

The study by Bilanges and Stokoe demonstrated that both RNAi and ASO approaches can reduce PDK1 expression, but the two modalities produced distinct transcriptional responses, consistent with their different mechanisms of target recognition and RNA silencing. These findings highlight that the choice between ASO- and RNAi-based approaches may influence both the magnitude and broader biological consequences of gene silencing. [7, 36, 38–42, 49, 96].

Overview of In silico tools for ASO design

Drug design has benefited over time using advanced computerized modelling tools and procedures that have been created specifically for the area of ADMET prediction using virtual screening and molecular docking. These tools have made great strides forward but are considerably less effective when applied to nucleic acid-based therapies such as ASOs. The virtual tools that exist for nucleic acid-based therapy (ASOs) have great potential as a growing area; however, they are significantly less comprehensive than those for small molecules [43, 44]. Currently available tools have thus considered several important variables such as oligonucleotide specificity, thermodynamics and design workflows that small molecule developers have at their disposal. Some tools, such as Sfold and PFRED, have utilised RNA secondary structures and accessibility to improve ASO binding, while others such as ASOptimizer have focused on troubleshooting and minimizing off-target effects of ASOs. Despite these advances, the In-silico tools employed for ASOs face limitations, particularly regarding complex RNA structure and prediction of long-term stability [45–47].

There are many benefits to the use of In-silico tools for the design of ASOs, including performing rapid screens of hundreds of target sequences and assessing critical parameters such as melting temperature, hybridization efficiency, and off-target effects. In-silico tools therefore minimize the reliance on experimental approaches that are more expensive and labour-intensive. There are several options available that integrate parameters of ASOs [e.g. Sfold and ASOptimizer], thus allowing for assessment via In-silico methods to increase ASO potency and specificity. Nevertheless, as mentioned earlier, limitations exist in the incomplete modeling of RNA dynamics and the number of available tools [45, 47].

The addition of advanced bioinformatics pipelines adds another level of complexity to ASO design by providing efficient workflows that generate and evaluate multiple ASO candidates through high-throughput workflows so that researchers can concentrate their efforts on the best candidates. Many of these bioinformatics pipelines will also provide off-target predictions and improve binding specificity through deep learning algorithms to limit any adverse effects on RNA sequences of non-targets. Most importantly, In-silico workflows increase the speed of developing safer and more effective ASO products; however, continued development is necessary to reach the full use of their potential in therapeutic applications [43, 45] (Fig. 3).

Fig. 3.

Influence of RNA secondary structure on ASO accessibility and structure-aware targeting

Key features and considerations for ASO design

Designing effective ASOs depends on several important parameters, such as optimizing GC content for stability, having strong binding energy to inhibit target mRNA, and minimizing off-target effects that may be caused by cross-reactivity. Design parameters such as oligo length, thermodynamic parameters and primary and secondary structures can enhance binding efficacy. Other important parameters for maximizing therapeutic potential include RNA stability, delivery strategies, immunogenicity, and stability in vivo. By incorporating these parameters of ASOs, it is possible to achieve specific targeted silencing of gene expression with limited off-target effects in a variety of research and therapeutic applications.

• Size of the Gene: The targeted gene’s size plays a role in designing the ASO. Bigger genes might have more possible binding sites and therefore could complicate the process of selecting the best target sequence. The targeted region in the gene (e.g., exon, intron, 5’-UTR, 3’-UTR) also matters because it could additionally affect the efficacy of the ASO. For instance, the ASO could be a more effective translational initiation blocker by targeting the start codon region instead of adding it to the 3’-UTR region not since it is in such a distance from the start codon. Plus targeting splicing sites would affect mRNA processing [30].

• Length of the Oligo: lengths of ASO normally range from 15 to 25 nucleotides but can range from 13 to 40 nucleotides as a function of the exact application. Short oligos may exhibit less specificity and less stable binding affinity to target mRNA, while longer oligos will be more specific and stable but carry a greater risk of off-target effects and may be more challenging to deliver to cells. The optimal length achieves a balance among stability [binding affinity], specificity, and synthetic feasibility [9, 37].

• GC Content (40% ≤ GC % ≤ 60%): An ideal GC content of between 40 and 60% strengthens the balance between stability and flexibility within a higher order, DNA-RNA hybrid. A GC Watson Crick base pair has three hydrogen bonds, whereas the AT [adenine–thymine] base pair has only two hydrogen bonds; therefore, GC rich regions provide a more stable hybridization. However, high GC content can increase binding affinity, which may lead to non-specific binding, or difficulty in disrupting the binding site. Similar conclusions can be drawn for low GC content in terms of instability, as a high number of AT base pairs in the oligo-mRNA hybrid may decrease stability and lead to a failure to silence the mRNA [9, 42].

• Target Site Selection: The exact spot on the mRNA that the ASO targets can provide another way in which its efficacy can vary. For example, targeting the 5’-UTR or start codon can inhibit the initiation of translation, while targeting splice sites can affect the processing of mRNAs. Careful consideration of which site to target can increase the potential for the oligo to be maximally functional. The accessibility of the region of the target mRNA is also critical. Some regions of the mRNA may be occluded by proteins or secondary structures, keeping them from being as accessible to oligo binding. Making selections for target regions known to be accessible and exposed increases the likelihood of binding [9].

• ASO Binding Energy (≤ −8 kcal/mol): The amount of energy needed for an ASO to first bind with its target mRNA describes the ASO’s affinity for and interaction with its target mRNA, as indicated by a binding energy of ≤ −8 kcal/mole, which is healthy for a molecule to bind with its target mRNA normally. The ability of an ASO to interact strongly with a target mRNA will allow the molecule to effectively reduce or inhibit the translation of the target mRNA by denying ribosomes the ability to translate the target mRNA or through the enzymatic removal of the target mRNA via RNase H activity. Conversely, a binding energy that is more positive suggests a weak interaction between an ASO and mRNA, and likely to produce an ineffective gene silencing effect [33, 34].

• Minimum Cross-Reactivity with Other Gene Transcripts: Specificity plays a major role in ASO design. The oligo has to bind selectively to the target mRNA without reacting with other non-target mRNA transcripts. Cross-reactivity can contribute to off-target effects, and such effects could cause unintended silencing of a gene and the resulting side effects. To reduce this risk, the oligo sequence must be well-designed such that it is minimally homologous with other sequences in the genome, especially in the 3’-untranslated regions (3’-UTRs), which are the target of oligos. They can use sophisticated bioinformatics tools to forecast the possible off-target effects by aligning the oligo sequence with the whole genome. This assists in reducing unwanted interactions with non-target transcripts or genes and lowering the risk of side effects [31, 33].

• Highest Oligo Binding Energy and Low Binding Site Disruption Energy: This parameter focuses on the importance of finding oligos that bind tightly to the target site (maximum binding energy) but in such a way that the binding does not severely disturb the natural conformation of the mRNA or form unwanted secondary structures. A low disruption energy guarantees that the binding site is still accessible and functional, enabling the oligo to bind efficiently without interference from secondary structures such as hairpins or loops that may occur in the mRNA [34, 36].

• Secondary Structure: The mRNA targeted by the ASO and the ASO itself both exhibit a variety of complex RNA secondary structures, including hairpin structures, loops, and stems. Such structural complexity can present challenges for ASO binding to the mRNA. Therefore, it is beneficial to ensure that the design takes this complexity into account to maximize the likelihood that an ASO will access its binding site on the target mRNA. In addition to avoiding regions of stable secondary structure formation on the mRNA, the ASO should be developed to exhibit the least amount of potential for secondary structure formation since secondary structure formation has the potential to diminish the affinity and/or specificity of the ASO for its target [33, 34, 36].

• Thermodynamics: Designing oligonucleotides requires considering thermodynamic characteristics such as melting temperature ™, which is defined as the temperature at which half of an oligonucleotide-mRNA pair will undergo denaturation. A high Tm indicates that the oligonucleotide-mRNA interaction will remain intact when exposed to physiological temperature yet will open again under exposure to physiological conditions. The degree of thermodynamic stability of an oligonucleotide influences its ability to prevent non-specific interactions with irrelevant mRNA sequences and to form secondary structures that can impact the performance of the oligonucleotide [33, 34].

• Chemical Modifications: Such changes enhance the stability and binding affinity of the ASOs through protection against nucleases, and can provide enhanced pharmacokinetics, thereby reduce immune activation and augment the oligo’s in vivo efficacy [13, 25].

• Nuclease Resistance: Degradation of ASOs may occur through their interaction with nucleases located in biological systems, and so a comprehensive strategy of ASO chemical modification combined with an ASO sequence design approach may aid in enhancing the stability of the ASO within a biological environment, resulting in greater efficacy against target genes.

• Delivery Method or Cellular Uptake Efficiency: The ASO method is only as effective as the delivery method used. The delivery methods available offer a variety of ways to increase the efficiency and distribution of an oligonucleotide, thus increasing its potential as a therapeutic agent. Examples include nanoparticles, liposomes, and conjugating with cell-penetrating peptides (CPPs) [19, 29].

• Immunogenicity: ASOs may exhibit enhanced immunogenicity in some situations, especially with oligos containing unmethylated CpG motifs. Therefore, although developing a completely immunologically inactive oligo is likely impossible, by designing intentionally and making modifications an oligo can be transformed into a safe and reliable element for therapeutic use with reduced immunogenicity.

• Dose–Response Relationship: A dose–response relationship analysis of ASOs can help in Pinpointing those concentrations which can effectively shut down genes, while still controlling for toxicity. A therapeutic index and the determination of minimal effective dose can be determined through this method [30, 38].

• The design of ASOs involves optimizing numerous factors to achieve efficacy and safety. The most important factors include tuning the GC content for stability, achieving sufficient binding energy to inhibit mRNA, and minimizing cross-reactivity to limit off-target effects. Key considerations for effect binding include oligonucleotide length, thermodynamic properties and avoiding secondary structures. Other factors, such as increasing resistance to nuclease degradation, delivery systems, and minimizing immunogenicity, and ensuring in vivo stability are also very important for the maximization of ASO therapeutic potential. Altogether, these factors support the development of very specific and effective gene-silencing tools for research and therapeutic applications.

Key tools for ASO design

Different ASO design tools cater to specific needs based on the RNA target type, design complexity, and optimization requirements as mentioned Table 1. Sfold is ideal for researchers focused on RNA secondary structure accessibility, making it particularly useful for general ASO design across various RNA types, though it may lack detail for tertiary structures [47]. lncASO is specifically designed for long non-coding RNAs (lncRNAs), employing machine learning to consider oligonucleotide affinity and target dimerization, and thus is the optimal option for those interested in lncRNA targets. PFRED is notable for its ease of use and thus is ideal for those with less computational design experience but still permitting more advanced users to maximize ASO sequences with multiple design parameters such as stability and potency [46]. OptiRNAi targets siRNA design, which is most useful for researchers involved in gene silencing through RNA interference, although it does not always take secondary RNA structures into account [48]. siExplorer is more suited for siRNA than ASO design, but it combines different rules to maximize target accessibility and stability, so it is a good choice for researchers interested in siRNA [50]. Lastly, ASOptimizer, with its powerful machine-learning algorithms, is perfect for researchers who require highly optimized ASOs with minimized cytotoxicity, particularly when working with mRNA sequences and looking to optimize efficacy and safety [45].

Table 1.

The table provides a comparative overview of computational tools used for the design and analysis of ASOs and RNA-targeting therapeutics

Tool	Working principles	Advantage	Limitations	References	Link to the tool
Sfold	Sfold uses a Boltzmann ensemble approach to predict RNA secondary structures, sampling multiple conformations to assess the accessibility of target sites. It incorporates recent Turner free energy rules to evaluate the stability of structures	Evaluates multiple RNA conformations for better target accessibility prediction	- May not account for all possible modifications or secondary structures of target RNA - Limited to RNA sequences without extensive datasets for all potential targets	[47, 102]	https://sfold.wadsworth.org/cgi-bin/index.pl
PFRED	PFRED utilizes a user-friendly interface to allow scientists to design ASOs targeting specific genes. It incorporates multiple design parameters, such as stability, potency, and potential off-target effects, to optimize oligonucleotide sequences	Simplifies ASO design through an accessible interface and multiple design parameters	- May not adequately address all potential off-target effects - Complexity may deter less experienced users	[46]	https://github.com/pfred/pfred-gui/releases/tag/v1.0
OptiRNAi	OptiRNAi employs a set of design rules derived from empirical studies to predict effective target sequences for siRNA production, evaluating parameters like GC content and target accessibility	Empirically derived design rules improve target sequence selection for siRNA production	- May not account for secondary structures of target RNA adequately - Success rates can be low in practical applications	[48]	https://bip.weizmann.ac.il/toolbox/target/rna/siRNA.html
siExplorer	siExplorer analyzes target RNA sequences to identify optimal siRNA candidates by evaluating features such as target accessibility, off-target potential, and thermodynamic stability, integrating various design rules	Incorporates a comprehensive range of design rules for siRNA candidate evaluation	- Primarily focused on siRNA rather than ASOs - May not generalize well to all target types	[49]	https://rna.chem.t.u-tokyo.ac.jp/cgi/siexplorer.htm
ASOptimizer	ASOptimizer employs a two-stage approach to optimize ASOs targeting specific mRNA sequences, using a linear factor model and a deep graph neural network to refine ASO sequences for improved efficacy and reduced cytotoxicity	Incorporates machine learning models for more accurate ASO design, targeting both efficacy and cytotoxicity	- May not be effective for targets not represented in the training data - Requires robust datasets for training to ensure predictive accuracy	[45]	https://github.com/Spidercores/ASOptimizer

Structure prediction

The process of predicting the 3D structure of single-stranded DNA (ssDNA) oligonucleotides presents several challenges because of their varied uses in many different areas such as biosensors, DNA vaccines, and ASOs. The existence of many different conformational forms of ssDNA, including hairpin loops, bulge loops, and G-quadruplexes, has a major impact on how these molecules function in a biological context. Whereas extensive experimental data as well as the existence of well-defined models allows for effective prediction of protein structures, prediction of the three-dimensional structure of ssDNA has yet to be realized because of the lack of dedicated computational tools. For example, in contrast to there being specific folding pathways and structural motifs that characterize a protein, ssDNA has a much wider range of potential for structural variation, thus making predicting the 3D structure of ssDNA extremely complex [43, 44].

The previously proposed workflow by Jeddi and Saiz has attempted to bridge this gap by integrating existing RNA three-dimensional (3D) structure prediction tools and modifying them to ssDNA. Their strategy entails first predicting DNA secondary structure, predicting the 3D structure of ssRNA, and subsequently transcribing it into ssDNA, adding another level of complexity because precise conversion of the ribose into deoxyribose and the uracil into thymine is required [51]. The list of available softwares for prediction of the 3D structure of ssDNA are mentioned in Table 2.

Table 2.

This table gives an overview of software packages employed for molecular biology analyses, with special emphasis on structural predictions, molecular docking, and MD simulations. Tools are included for RNA/ssDNA secondary structure prediction, RNA tertiary structure, molecular docking, MD simulations, and.pdb file editing, as well as links of direct access

Name of the software	Use	Link to the tool	Reference
Mfold	ssDNA secondary structure prediction	http://www.unafold.org/mfold/applications/dna-folding-form.php	[52]
RNAfold	RNA secondary structure prediction	http://rna.tbi.univie.ac.at/cgi-bin/RNAWebSuite/RNAfold.cgi	[53]
Centroidfold	RNA secondary structure prediction	http://rtools.cbrc.jp/centroidfold/	[54]
RNA structure	Both RNA & ssDNA secondary structure prediction	http://rna.urmc.rochester.edu/RNAstructureWeb/Servers/Predict1/Predict1.html	[55]
KineFold	Both RNA & ssDNA secondary structure prediction	http://kinefold.curie.fr/cgi-bin/form.pl	[56]
RNA composer	RNA Tertiary structure	https://rnacomposer.cs.put.poznan.pl/	[57]
Pdb editor jar	.pdb file editor	https://www.bioinformatics.org/pdbeditor/wiki/	[61]
RNA2DNA	Three-dimensional structure of ssRNA to ssDNA conversion tool	https://github.com/krishnakumarbhat/RNAtoDNA https://github.com/AbhignaNagaraj/RNA2DNA/blob/main/RNA2DNA.py	[62, 63]
HNA DOCK	Molecular Docking	http://huanglab.phys.hust.edu.cn/hnadock/	[64]
H DOCK	Molecular Docking	http://hdock.phys.hust.edu.cn/	[65]
HADDOCK	Molecular Docking	https://rascar.science.uu.nl/haddock2.4/	[66]
Schrodinger-PIPER	Molecular Docking	https://www.schrodinger.com/platform/products/piper/	[67]
Schrodinger-DESMOND	MD simulations	https://www.schrodinger.com/platform/products/desmond/	[68]
Gromacs	MD simulations	https://www.gromacs.org/	[69]

Also, the metal ion-influenced polyanionic character of DNA makes the prediction of correct 3D structures even more challenging. Yet reliable prediction is a long-term goal, especially for bigger and more complicated ssDNA molecules, calling for ongoing research and development of dedicated tools and methods in this field.

Secondary structure prediction tools

Several computational tools are available for predicting nucleic-acid secondary structure, and their usefulness depends on the type of structural information required. Mfold uses thermodynamic optimization to predict possible folding patterns and has been widely used for nucleic-acid structure analysis [52]. RNAfold, available through the Vienna RNA WebSuite, uses minimum-free-energy (MFE) prediction and provides additional options for structural analysis [53]. CentroidFold uses a centroid-based approach, providing an alternative to conventional MFE prediction [54]. RNAstructure includes several algorithms for RNA secondary-structure prediction and related structural analyses [55]. KineFold differs from these approaches by allowing investigation of folding pathways and structural features such as pseudoknots and knots in RNA and DNA [56]. Thus, these tools offer complementary approaches, and the choice of method depends on the nucleic-acid type, sequence characteristics, structural features, and the specific purpose of the analysis. RNA secondary structure can also affect the accessibility of an ASO target site. Nucleotides located within stable stems or hairpins may be less readily available for hybridization, whereas loops, bulges, and other exposed regions can provide more accessible sites. For this reason, RNA structural accessibility should be considered together with sequence-based criteria when selecting ASO target sites. In addition to conventional duplex formation, alternative structure-guided approaches have also been explored. Triplex- forming peptide nucleic acids (PNAs) have been investigated for sequence- specific recognition of complex RNA structures [100]. A related dual-affinity approach uses triplex formation together with an adjacent duplex interaction to facilitate recognition of structured RNA targets [101].

Tertiary structure prediction

RNA Composer is a tool that can predict the 3D structure of large RNA molecules. It uses an automated approach to compose the 3D structure from secondary structure information. RNA Composer has been shown to produce accurate predictions for large RNA molecules [57].

Three-dimensional structure conversion of ssRNA to ssDNA:

To convert ssRNA 3D structures to ssDNA, tools like PyMOL and ChimeraX allow manual editing of nucleotides and sugar backbones, while pNAB and 3DNA facilitate structural rebuilding [58–60, 62]. Automated base conversion can be done with custom scripts in BioPython, which offer more flexibility for comprehensive transformations. RNA2DNA, the GitHub repository RNAtoDNA, which is developed by our team. Although this tool is not yet published and is protected under copyright, it is available for use on GitHub. It provides a Python-based tool designed to convert RNA sequences into DNA sequences. The script is specifically developed to automate the conversion of RNA 3D structures to their corresponding DNA forms by substituting uracil (U) with thymine (T) and adjusting the sugar backbone from ribose to deoxyribose. This is particularly useful for researchers working with nucleic acid modeling and in scenarios where they need to convert the three-dimensional structure of RNA data for DNA-based applications [61].

Molecular docking tools

Molecular docking approaches such as HADDOCK, HDOCK, and HNADOCK can be used to generate structural hypotheses for nucleic acid–protein interactions. However, their application to ASO–protein complexes remains limited by the conformational flexibility of single-stranded oligonucleotides, chemical modifications, protein conformational changes, and the limited availability of experimentally resolved ASO–protein structures. Therefore, docking-derived poses and scores should be considered hypothesis-generating rather than definitive evidence of ASO–protein binding. Experimental approaches, including biochemical binding assays, crosslinking/mass spectrometry, or high-resolution structural methods, are required to validate predicted interactions. [63–67].

Molecular dynamics simulations

Schrodinger-DESMOND and Gromacs are leading software tools for carrying out MD simulations and enabling scientists to investigate the time-dependent dynamic motion of ASOs and their bindings with target molecules. Through the simulations, knowledge about the ASO’s stability and conformational fluctuations under biological conditions can be obtained, which is pivotal in their optimized design and making predictions of behavior in vivo [68, 69].

Advancing ASO development through integrated in-silico and in-vitro approaches

In-silico approaches are now unavoidable in ASO design and optimization, providing tremendous cost, speed, and accuracy benefits. In silico tools enable high-throughput screening, high-speed simulation of RNA-oligo binding affinities, and prediction of off-target activity, facilitating the early stages of candidate selection. Additionally, In-silico models can predict the thermodynamic stability of oligonucleotide conformations, which is useful for designing ASOs. Although computational methods for predicting ASO activity are powerful, they will never fully capture the full extent of ASO complexity in the biological system. Therefore, an optimally combined approach of computational predictions followed by the experimental methods will allow for the direct evaluation of ASO efficacy, selectivity, and safety [70, 71]. The combination of In-silico methods with other preclinical experimental methods such as cell-based assays, exon skipping, and in vitro transcription provides an opportunity to validate the ASOs against the previously raised theoretical concerns. This combination provides an additional set of tools for the selection and modification of the oligonucleotide sequence to yield optimal ASOs. The addition of high-throughput sequencing and RNA immunoprecipitation also contributes to validating ASO/RNA target interactions and ensures the proper performance of ASOs within the biological complexities of cellular environments. Finally, by combining In silico and In vivo methods, researchers will be able to focus on a smaller design window and streamline the development of more effective therapies and potentially have an impact on the development of personalized medicines and the future evolution of drugs [71–73].

Gaps in current in-silico tools for ASO development

While promising, today’s In-silico technologies reveal a major limitation regarding the secondary and tertiary structures of ASOs as they relate to oligonucleotide stability and function. In addition, in vitro and in vivo testing has shown that current in silico predictive methods cannot realistically reproduce ASO secondary and tertiary structure, especially in a cellular context [77, 79]. Some of the research gaps include:

1. The limitations of current prediction algorithms on non-canonical structure: Current prediction methods for ASO prediction largely account for canonical Watson–Crick base-pairing interactions, but there is a substantial decline in predictive accuracy with respect to any non-canonical base-pairing structures (e.g., pseudoknot, G-quadruplex, and bulge-like) which play an important role in ASO activity via effective binding affinity or specificity. Existing prediction algorithms typically incorporate canonical base-pairing interactions in predicting ASO structure but do not account for non-canonical base-pair interactions, resulting in oversimplified predictions of ASO behavior in vivo [80, 81].

2. Limitations of RNA–Protein Interaction Simulation: In the past, although ASOs worked in an extremely dynamic system, most of these interactions with RNA-binding proteins were represented as static interactions in computational models. The proteins that bind RNA (RNA-binding proteins) and assist/translocate RNA (chaperone proteins) also have a relatively minor role in determining the precise shape and relative stability of the RNA molecule via post-transcriptional modifications. Most In-silico models have a very narrow perspective and only consider the relationships between individual RNA molecules within the limited molecular environment of ASOs. The inability to predict the function of ASOs in living organisms substantially constrains the ability to accurately predict the mechanism by which RNA-binding proteins interact with ASOs [80–82].

3. Environmental effects at the cellular level: Currently available In-silico models do not adequately account for environmental factors that may affect the configuration and behavior of ASOs, such as the local concentration of RNA and ionic conditions, as well as endogenous RNAs competing with ASOs. As such, these environmental effects make it difficult to predict the actual behavior of ASOs when designed using only existing In-silico models when not coupled to their actual cellular environment [83, 84].

Experimental data for refining predictive models

High-resolution experimental information will be a key component to filling the current research gaps in ASO design. Structural information for ASOs in biologically complex situations can be obtained through cutting-edge techniques such as cryo-EM, X-ray crystallography, and SAXS, which help to refine computational constructs (i.e., to improve predictive secondary/tertiary structural information as well as exploring RNA–protein interactions). Crosslinking and mass spectrometry provide additional useful information about the interaction of ASOs with the proteins necessary for RNA processing and degradation. The combination of these two experimental methods with computational modeling will enhance ASO prediction accuracy through the development of hybrid data models that incorporate computer-derived molecular dynamics simulation results and experimental structural information. Notably, these methodologies aid in the accuracy of prediction; however, they also contribute to increased costs and timelines related to the development of ASOs due to the costs and time-consuming nature of structural studies like cryo-EM and mass spectrometry as a balancing act between the accuracy of research and costs is always important in ASO development [79, 85].

Addressing the cost and complexity of ASO development

Incorporating high-resolution experimental data into computational models is very important to advance ASO design, though the associated development costs may vastly increase. Advanced approaches such as cryo-EM, X-ray crystallography, and mass spectrometry all have significant expense, require specialized laboratory space, and can be time-consuming, creating a burden for researchers, especially in my niche ASO approaches. While all these aspects are expensive and labor-intensive, the increased accuracy and efficacy of their use will provide value, especially in therapeutic contexts. While In-silico tools can be a fast and cost-effective means of screening compounds or performing structural calculations, the nuanced aspects of RNA structure cannot be predicted by them. Combining these approaches, ASO predictions become more reliable, advancing the development of precise, targeted therapeutics in personalized medicine [86–88].

Delivery challenges and strategies

The effective delivery of ASOs to target tissues remains a challenge. Various strategies are being explored to enhance delivery, including:

Chemical modifications to improve stability and cellular uptake

To improve ASO stability and/or binding affinity and/or optimal therapeutic efficacy, ASOs have undergone dramatic changes that involve numerous distinctions in the chemical structure of ASOs. The most common modifications of ASOs are represented as the three generations of ASOs, beginning with: 1st Generation (Backbone Modification)—The 1st generation has been focused primarily on the modification of the [d]phosphodiester backbone. One of the modifications that stands out among others is that of phosphorothioates (PS). Phosphorothioate contains a sulfate atom in the phosphodiester backbone that replaces the central O atom in the phosphate group. PS-modified ASOs exhibit higher nuclease resistance, longer plasma half-life, and enhanced binding to plasma proteins. However, they can also interact with various cellular proteins and components of the innate immune system, leading to potential pro-inflammatory effects and side effects such as fever, thrombocytopenia, and leukopenia [75, 78, 89–91].

Second Generation (Glycosyl Modification): Building upon the backbone modifications, second-generation ASOs incorporate sugar modifications, such as 2’-O-methyl (2’-Ome) and 2’-O-methoxyethyl (2’-MOE). These modifications further enhance nuclease resistance, binding affinity, and tissue uptake. Chimeric 'gapmer' ASOs, consisting of a central ‘gap’ region containing DNA or PS-DNA flanked by 2’-OMe or 2’-MOE-modified nucleotides, combine the RNase H-mediated cleavage of the central DNA region with the nuclease resistance and steric interference of translation provided by the modified ends. While safer than first-generation ASOs, a subset of 2’-MOE-modified ASOs can still induce proinflammatory cytokines and interact with innate immune receptors [70, 76, 92] (Fig. 4).

Fig. 4.

Chemical modifications of antisense oligonucleotides

Third Generation (Other Modification): The third generation of ASOs includes modifications such as locked nucleic acids and phosphorodiamide morpholino oligonucleotides (PMOs). LNAs are chemically modified nucleotides with a methylene bridge between the 2’ oxygen and 4’ carbon of the ribose, which improves binding affinity and nuclease resistance. However, LNAs do not activate RNase H and require incorporation into chimeric oligonucleotides to restore RNase H-mediated cleavage. PMOs are neutral ASOs with a morpholino ring substituting the pentose sugar and phosphoramidate bonds instead of phosphodiester bonds. PMOs exert their mechanism of action through steric interference with ribosomal assembly, leading to translational arrest. They exhibit fewer nonspecific properties and lower toxicity compared to PS-modified ASOs but have reduced cellular uptake, which can be improved by conjugation with peptides.

These advancements in chemical modifications have significantly contributed to the development of more potent and specific ASO therapeutics, with several FDA-approved drugs and many more in clinical trials [94, 95].

Conjugation with targeting ligands or cell-penetrating peptides (CPPs)

Successful administration of ASOs is an important consideration in therapy, depending on target localization and delivery means such as intravascular or subcutaneous injection, lipid nanoparticles, and conjugation with targeting ligands or CPPs. CPPs are short peptides capable of facilitating cellular uptake of cargo molecules, including nucleic acids. Because ASOs may undergo degradation or show limited cellular uptake in biological environments, delivery systems such as liposomes and nanoparticles can provide protection and enhance cellular entry. Targeting ligands such as GalNAc can further improve tissue-selective delivery, particularly to hepatocytes. Advances in conjugation and delivery strategies may therefore improve the tissue distribution, cellular uptake, and therapeutic potential of ASOs. Because ASOs may not be structurally stable in biological environments, delivery frameworks such as liposomes and nanoparticles also provide a protective layer around ASOs, enhancing uptake. Alternative modes of conjugation are using targeting ligands, like GalNAc for liver cells. The implementation of Computational Prediction Programs will increase the specific and enhanced delivery of ASOs to these areas of the body while reducing any unacceptable side effects. Advances in the design and functional capabilities of ASO delivery approaches can be instrumental in enhancing the total therapeutic potential of ASOs for all types of gene regulation [97, 98].

Formulation with nanoparticles or lipid-based carriers

Lipid-based carriers and nanoparticles are innovative strategies that can help to protect against degradation and optimize pharmacokinetics to improve ASO delivery and efficacy. Lipid-based nanoparticles (LNPs) containing ionizable cationic lipids, phospholipids, polyethylene glycol (PEG) lipids and cholesterol protect and stabilize ASOs upon encapsulation and prolong circulation time. After the LNP is delivered to its target cell, it will fuse with the cell’s membrane to release the ASOs into the cytoplasm and subsequently allow them to perform their therapeutic effect. Other nanoparticles or formulations aside from LNPs have been studied, including polymeric nanoparticles, dendrimers, and metallic nanoparticles. Many polymeric nanoparticles are made from PLGA (poly (lactic-co-glycolic acid)) and PEI (polyethyleneimine) so that they protect the active ingredient from degradation, while dendrimer-based nanoparticles have been studied for their ability to facilitate cellular uptake by virtue of electrostatic interactions. Metallic nanoparticles, such as gold and iron oxide nanoparticles, also provide unique properties for targeted delivery. These newer formulation technologies can significantly improve ASO stability and cellular uptake and targeted delivery. However, matching efficacy through effective, safe, and targeted nonviral delivery systems, remains an important emerging area of research to achieve the full therapeutic potential of ASOs. To solve the persistent issues associated with ASO delivery and efficacy, continued innovations in the development of novel nanomaterials and carrier technologies will lead to next-generation ASO-based therapies [74].

FDA-approved ASOs

ASOs are an innovative class of drugs that have made substantial strides in the past few years with multiple FDA approvals indicating their therapeutic potential for different genetic disorders. The list of major brands and their type of treatments are displayed in Table 3. The first ASO approved by the FDA was Vitravene (fomivirsen) in 1998 for the treatment of cytomegalovirus retinitis. It was followed by Kynamro (mipomersen), approved in 2013, targeting apolipoprotein B mRNA for the treatment of homozygous familial hypercholesterolemia. The approval of these products has highlighted the therapeutic effectiveness of ASOs in treating orphan conditions that have insufficient therapeutic options, demonstrating their capability to interfere directly with disease-producing genes using a new mechanism of action. The increase in the use of chemical modifications to ASOs to increase their stability, binding and pharmacokinetics is the basis for their efficacy in the treatment of disease. Two important chemical modifications of ASOs that are key to their ability to resist nucleolytic degradation and off-target toxicity are phosphorothioate backbones and 2’-O-methyl (2’-Ome) substitutions. The recent approvals of Inotersen (Tegsedi) and Spinraza (Nusinersen) are indicators of continued growth in the application of ASOs to treat conditions such as hereditary transthyretin-mediated amyloidosis and spinal muscular atrophy, and potentially to treat other medical conditions beyond genetic disorders, including in the areas of oncology and infectious diseases, where ASOs may also be useful. The development of splice-switching oligonucleotides, e.g., Eteplirsen and Golodirsen, are an example of the use of ASOs to specifically target exon-specific splicing of the dystrophin gene, thus enabling the production of functional dystrophin protein in boys with DMD. The continuing developments in ASO technology, such as ASOs targeting multiple exons of the dystrophin gene, suggest that unmet medical needs associated with genetic diseases continue to be addressed using genetic medicines that can effectively provide intervention for these diseases. Overall, the approval of ASOs by the FDA solidifies their potential to function as therapeutic agents, offering new opportunities to improve drug delivery and utilize the advancements in oligonucleotide chemistry [20, 35, 93, 99].

Table 3.

This table summarizes ASO therapies, detailing brand names, treatment types, targets, target organs, chemical modifications, administration routes, diseases, approval years, status, and references

Brand name	Type of treatment	Target	Target organ	Chemical modifications	Route of administration	Disease	Year of approval	Status	ASO delivery technique
Vitravene (Fomivirsen)	ASO	mRNA encoding IE2	Eye	2’-H	Intravitreal	CMV retinitis, HIV Infections	1998	Completed	Local intravitreal injection (naked ASO)
Kynamro (Mipomersen)	ASO	ApoB-100 mRNA	Liver	2’-MOE	Subcutaneous	Homozygous familial hypercholesterolemia	2013	Completed	Systemic phosphorothioate ASO delivered by subcutaneous injection
Tegsedi, AKCEA-TTR-LRx (Inotersen)	ASO	Hepatic transthyretin [TTR] mRNA	Liver	2’-MOE	Subcutaneous	Hereditary ATTR Amyloidosis [hATTR]	2018	Completed	Systemic phosphorothioate ASO delivered by subcutaneous injection
Exondys 51 (Eteplirsen)	Splice Switching Oligo	DMD 001-gene [exon 51 target site]	Muscle	2′-MOE, PMO	Intravenous	Duchenne muscular dystrophy	2016	Completed	PMO-mediated systemic intravenous delivery
Vyondys 53 (Golodirsen)	Splice Switching Oligo	DMD pre-mRNA splicing [exon 53 skipping]	Muscle	2′-MOE, PMO	Intravenous	Duchenne muscular dystrophy		Recruiting	PMO-mediated systemic intravenous delivery
Viltepso (Viltolarsen)	Splice Switching Oligo	DMD pre-mRNA splicing [exon 53 skipping]	Muscle	2′-MOE, PMO	Intravenous	Duchenne muscular dystrophy		Recruiting	PMO-mediated systemic intravenous delivery
Amondys 45 (Casimersen)	Splice Switching Oligo	DMD pre-mRNA splicing [exon 45 skipping]	Muscle	PMO	Intravenous	Duchenne muscular dystrophy		Recruiting	PMO-mediated systemic intravenous delivery
Spinraza [Nusinersen]	Splice Switching Oligo	Survival of motor neuron 2[SMN2] pre-mRNA splicing [exon 7 inclusion]	Central Nervous System	2′-MOE, Fully modified	Intrathecal	Type 1, 2, and 3 spinal muscular atrophy	2016	Completed	Direct intrathecal CNS delivery

Conclusion

Advances in ASO technology, especially in association with the In-silico approaches, have changed the landscape of molecular medicine. ASOs have emerged as a robust option for precise modulation of genes, targeting applications in a wide range of genetic disorders and therapeutic approaches.

By utilizing computational resources that enable the design of ASOs that are targeted to specific RNA sequences and aid in the optimization of binding affinity and specificity, as well as stability, ASO capabilities have rapidly expanded. The In-silico techniques not only help researchers in terms of designing ASOs, but they also assist in decreasing possible ‘off-target’ effects, which means they will likely have an increased safety profile and increased effectiveness. The rational design approach will enhance understanding of the physicochemical properties, RNA interactions and structural dynamics of ASO, thereby improving the ability to predict their behavior in biological systems. Future research should include optimizing ASO design and enhancing delivery and therapeutic efficiency.

The clinical success of ASOs supports their continued development and broader application in personalized medicine. Future efforts should focus on expanding the range of therapeutic targets while optimizing the therapeutic index through innovative ASO designs. Integrating next-generation computational approaches with ASO platforms will facilitate the development of targeted therapeutics for genetic diseases with reduced toxicity. In addition, challenges related to immunogenicity and delivery efficiency must be addressed as ASO-based therapies continue to expand into the treatment of cancer and rare diseases. The integration of experimental approaches with in silico modeling will accelerate the development of next-generation, highly specific ASOs with broader therapeutic potential.

Author contributions

Investigation- AN.; Writing original draft preparation- A.N., K.C.P., B.H., S.S., Reviewing and Editing- C.D., B.S., S.P.K., C.S. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Data availability

All the data originated from this research is available from the authors upon request.

Declarations

Competing interest

The authors declare that there is no competing interest.

Conflict of interest

The authors declare that the study was not conducted under any conflict of interest.