Antisense Oligonucleotide Design
Mechanism-specific ASO design that connects transcript biology, hybridization properties, specificity, and experimental decision-making.
Discuss Your ProjectOverview
Antisense oligonucleotides can recruit RNase H1 to degrade RNA, redirect pre-mRNA splicing, block translation or RNA–protein interactions, and modulate noncoding transcripts. These mechanisms require different target windows, architectures and experimental readouts. CD ComputaBio defines the intended mechanism first, then integrates transcript annotation, local RNA context, sequence properties, chemical-design constraints and hybridization-dependent off-target risk to prioritize an experimentally tractable ASO panel.
A candidate that is complementary to a gene is not automatically active, isoform-selective or safe. The relevant sequence may be absent from the disease isoform, inaccessible within RNA structure or ribonucleoprotein complexes, altered by a common variant, or shared with unintended transcripts. Conversely, a computational score cannot establish cellular uptake, endosomal escape, RNase H1 recruitment, splice correction or tolerability. Our service converts these uncertainties into explicit design filters and validation gates. It can be ordered as a focused module within our Bioinformatics Services portfolio without requiring synthesis or wet-lab work.
Research Questions and Inputs
| Input | Accepted information | Design consequence |
|---|---|---|
| Target and mechanism | Gene symbol plus stable transcript accession; desired knockdown, exon skipping/inclusion, translation blocking, start-site blocking, poison-exon modulation or noncoding-RNA interference. | Determines whether candidates target mature RNA, pre-mRNA, splice junctions, regulatory motifs or other defined windows. |
| Species and model | Human, mouse, rat, non-human primate or another organism with an adequate genome/transcript annotation; tissue, cell type, disease state and intended cross-species testing. | Defines transcript references, ortholog conservation, expression filters and surrogate-species options. |
| Sequence data | FASTA, GenBank, Ensembl/RefSeq accessions, genomic coordinates, custom constructs, patient variants or assembled transcripts. RNA-seq FASTQ/BAM, count matrices or junction counts may be accepted when isoform evidence is needed. | Anchors coordinate systems, target coverage, splice-event evidence and personalized designs. |
| Chemistry constraints | Requested oligo length, phosphorothioate pattern, DNA gap, 2'-MOE, 2'-O-methyl, other permitted chemically modified nucleotide building blocks, conjugation constraints and synthesis exclusions. | Changes affinity, nuclease stability, RNase H compatibility, motif interpretation and acceptable sequence-property ranges. |
| Metadata | Genome/annotation build, sample group, biological replicate, batch, donor, genotype, treatment, dose, time, library protocol and strandedness. | Prevents isoform and differential-expression conclusions from being confounded by unrecorded design variables. |
Common sequencing sources include Illumina bulk RNA-seq, targeted amplicon sequencing and long-read transcript data from PacBio or Oxford Nanopore. Long reads can resolve isoforms but may require platform-specific error handling; short reads provide depth but junction support and transcript quantification remain model-dependent. Client-supplied annotations are versioned and compared with the selected reference.
Core Services
Enumerate candidates across eligible exonic, intronic or noncoding regions, then evaluate central DNA-gap compatibility, modified-wing constraints, GC distribution, repeats, self-complementarity and sequence motifs. Architecture is reported separately from sequence so that chemistry assumptions remain traceable.
Map exon–intron boundaries, alternative junctions, branch-point or regulatory neighborhoods when evidence supports their use, and enumerate steric-blocking candidates across defined splice-control windows. Predicted exon effects remain hypotheses until confirmed by isoform-resolved assays.
Identify which coding and noncoding isoforms contain each site, whether a candidate is pan-transcript or isoform-selective, and whether alternative transcription starts, polyadenylation or exon usage changes coverage in the relevant tissue.
Combine local secondary-structure estimates, ensemble accessibility, base composition, homopolymers, repeat content and self-dimer or hairpin risk. Structural predictions are treated as relative evidence because cellular proteins and RNA dynamics can alter accessibility.
Search transcriptomes and, when relevant, pre-mRNA or genomic contexts for close complementary sites. We record mismatch number and position, transcript region, abundance and gene relevance, and intersect common or client-specific variants that may disrupt on-target binding or create alternative sites.
Balance target coverage, sequence diversity, mechanism-specific properties, predicted accessibility, specificity, conservation and synthesis constraints. Rather than returning many nearly identical oligos, we select a diverse panel that tests distinct windows and reduces correlated design failure.
Six-Stage Workflow
The workflow preserves reference releases, coordinate conversions, software, parameters, rejected motifs and ranking weights so designs can be reproduced or updated.
Confirm the desired RNA outcome, ASO class, target tissue, species, chemistry boundaries, validation assay and decision criteria.
Resolve gene and transcript identifiers, isoforms, exon coordinates, strand, variants, orthologs and relevant tissue-expression evidence.
Generate candidate sites within mechanism-appropriate regions while retaining coordinates, isoform coverage and sequence provenance.
Assess GC profile, motifs, repeats, self-structure, predicted accessibility, duplex properties and chemistry-specific restrictions.
Enumerate complementary sites, annotate mismatch patterns and expression context, and evaluate variant and cross-species consequences.
Rank candidates, examine threshold sensitivity, select nonredundant windows, and pair each risk with an experimental control.
Scientific Evidence
Open-source frameworks such as PFRED illustrate a reproducible design pattern: retrieve transcript and ortholog information, enumerate oligos, calculate properties, add efficacy and off-target annotations, then filter candidates. Our implementation is project-specific and does not imply that one public score is sufficient. The displayed OA figure shows how candidate generation becomes a selection problem after descriptors are calculated.
For splice-switching projects, experimentally curated models such as eSkip-Finder demonstrate that sequence and positional features can inform ranking. However, performance measured on known exon-skipping datasets may not transfer to a different gene, chemistry, species or assay. For gapmers, sequence complementarity, RNA accessibility, gap architecture and off-target hybridization are considered jointly, followed by concentration-response testing.
Statistical and AI Strategy
Computational enumeration does not require biological replicates, but model training and experimental confirmation do. Screening should include multiple ASOs per target region, chemistry-matched negative controls, positive controls where available, independent biological replicates and concentration ranges. Power is tied to assay variance and the effect size needed for progression.
Plate, synthesis lot, transfection, donor, passage, dose and time can confound ASO effects. Randomization and blocked designs are preferred. If all active candidates are tested in one batch and controls in another, statistical adjustment cannot recreate the missing comparison.
Primary endpoints—RNA reduction, exon inclusion/skipping, protein change or phenotype—should be prespecified. We report effect sizes and uncertainty intervals, not only P values. Screening across many oligos and transcripts requires false-discovery-rate or other appropriate multiplicity control.
When machine learning is used, related oligos, the same target gene and measurements from the same experiment must not leak across training and test sets. Feature selection, scaling and tuning stay within training folds; nested cross-validation and gene-held-out evaluation are preferred.
External validation should include new target genes, laboratories, chemistries or species when the intended use requires that level of generalization. We document the training domain for oligo length, chemistry, mechanism, assay and organism. Predictions outside that domain are flagged, model disagreement is retained as uncertainty, and rankings are subjected to threshold sensitivity analysis. AI scores do not convert an in silico candidate into a validated therapeutic lead.
Deliverables
Mechanism, transcript accessions, isoform map, target windows, genome build, orthologs and key assumptions.
ASO sequences, orientation, coordinates, length, target region, isoform coverage and proposed architecture.
GC profile, motifs, repeats, self-structure, relative accessibility, duplex features and filter outcomes.
Ranked complementary sites with mismatch patterns, transcript annotations, expression context, variants and cross-species matches.
Decision-ready shortlist with transparent scores, diversity across target windows, reasons for inclusion and residual uncertainties.
Recommended controls, assays, doses, time points, orthogonal endpoints, software versions, parameters and reference provenance.
Applications and Validation
Prioritize RNase H1-compatible gapmers for coding RNA, nuclear-retained RNA or selected noncoding targets, with transcript coverage and hybridization-dependent off-target review.
Design candidates for exon skipping, exon inclusion, cryptic splice suppression or poison-exon control, using junction-aware coordinates and isoform-specific readouts.
Evaluate allele-discriminating or personalized sites, common polymorphisms, haplotype uncertainty and conservation for preclinical species without assuming that a mismatch guarantees selectivity.
Recommended validation begins with multiple independently located ASOs, concentration–response curves and chemistry-matched controls in a relevant cell model. Gapmers can be assessed by RT-qPCR or digital PCR, followed by protein measurement and transcriptome-wide RNA-seq for selected leads. Splice-switching candidates require junction-specific RT-PCR, capillary electrophoresis or targeted sequencing to quantify intended and unintended isoforms. Reporter assays can isolate a target window, but should not replace endogenous-locus testing. RNA structure probing, RNase H cleavage mapping or orthogonal binding measurements may investigate mechanism. Rescue experiments or a second ASO producing the same molecular and phenotypic result can strengthen on-target interpretation. High-value leads should be retested in an independent experiment, disease-relevant cells and, where needed, a justified in vivo model.
Project Strategy and Boundaries
For a discovery panel, stable target accessions, species, mechanism and chemistry constraints may be sufficient. Isoform-selective or splice-correcting programs usually require transcript-level evidence and exact exon coordinates. Translational programs benefit from tissue expression, population variants, cross-species conservation and a prespecified plan for unbiased expression profiling. If the reference is incomplete or the target is poorly expressed in available data, the limitation is reported rather than hidden by a score.
Sequence design predicts compatibility and prioritizes experiments; it does not establish delivery, intracellular concentration, protein binding, immune activation, class-related toxicity or clinical efficacy. An observed expression change may be a downstream consequence of intended target modulation rather than direct hybridization. Candidate sites are therefore described as predicted or prioritized, and causal claims require direct or convergent experimental evidence. Chemistry recommendations are design hypotheses subject to synthesis feasibility, analytical quality control, formulation and toxicology review.
At project initiation, we define the progression rule—for example, a minimum molecular effect at a tolerated concentration, reproducibility across biological replicates, an acceptable off-target expression profile and confirmation with an independent ASO. This makes computational ranking and wet-lab evidence part of one auditable decision process.
References
Share your target transcript, species, intended mechanism, chemistry constraints and validation model.
Submit your project details below, and our team will respond within 24 hours.
Talk to our technical team about your project!
I Want To Talk