Use OligoWalk Online
Commercially Available OligoWalk No-Code Web Server
OligoWalk: siRNA Design and Silencing Efficacy Prediction Using Hybridization Thermodynamics
OligoWalk is an online small interfering RNA (siRNA) design tool that ranks candidate siRNAs using predicted hybridization thermodynamics. It was described by Zhi John Lu and David H. Mathews of the University of Rochester Medical Center in Nucleic Acids Research (2008, Web Server issue). Given a target mRNA sequence, the server lists siRNA candidates ranked by their probability of being efficient, defined as silencing efficacy greater than 70%. The server was freely accessible at rna.urmc.rochester.edu/servers/oligowalk when the paper was published.
Not every siRNA duplex silences its target, so the authors note that siRNA design involves a significant cost in trial and error. Empirical sequence rules, such as low G/C content, an A at position 3, no G or C at position 19, and asymmetric stability of the terminal base pairs, help but do not capture everything. The paper also highlights that self-structure of the target and of the oligonucleotide matters for effective binding, so a good candidate has an accessible binding site and a low-stability duplex.
Property | Detail |
|---|---|
Input | Target RNA sequence (A, U, T, G, C), up to 10,000 nucleotides |
Default siRNA | 19-nucleotide RNA oligonucleotide |
Classifier | Support vector machine trained on 2431 experimental results in human cells at 37 C |
Positive predictive value | 0.786 (78.6% of selected siRNAs are efficient) |
Output | Ranked siRNA table plus a thermodynamic details table |
Venue | Nucleic Acids Research 2008, 36, W104-W108 |
How OligoWalk Predicts Efficient siRNA
OligoWalk predicts free energy changes for an siRNA binding to its target, considering unimolecular and bimolecular self-structure of the siRNA and unimolecular self-structure of the target at the binding region. These structures are in equilibrium with each other and with the hybridized state. The thermodynamic terms plus local sequence features are fed to an embedded support vector machine, which outputs the probability that each candidate silences efficiently.
Scan the target: the server walks along the mRNA and evaluates each candidate oligonucleotide.
Prefilter: an optional filter based on the empirical rules of Reynolds et al. removes candidates scoring below six points before folding.
Fold and compute free energies: by default the target is folded and refolded for each candidate using a partition function calculation, the most rigorous mode.
Classify: the SVM ranks candidates by probability of efficacy greater than 70%.
Binding Mode and Structure Prediction Options
No target structure: fastest, but not recommended for accurate siRNA prediction.
Break local structure: the target structure is fixed and only base pairs at the binding site are broken.
Refold target: the default and most rigorous mode, with optimal structure, suboptimal structure (up to 1000 structures within 10% of the optimal free energy) or partition function methods.
Thermodynamic Outputs and Runtime Benchmarks
The details table reports terms in kcal/mol, which advanced users can use to understand why a candidate ranks well.
Output column | Meaning |
|---|---|
Overall | Overall free energy of oligonucleotide-target binding, affected by oligonucleotide concentration |
Duplex and Tm-Dup | Duplex free energy and melting temperature of the antisense-sense duplex |
Break-targ | Free energy cost of opening target base pairs at the binding site |
Intraoligo and Interoligo | Intramolecular and intermolecular oligonucleotide structure terms |
End_diff | Stability difference between the 5 prime and 3 prime ends; functional siRNAs prefer an unstable 5 prime end |
Prefilter_score | Score from the empirical Reynolds et al. rules |
With default options (19-nucleotide oligo, 800-nucleotide folding size, partition function, full scan, prefilter on), the paper reports run times from 57 minutes for a 730-nucleotide target to about 6 h 53 min for a 3460-nucleotide target on the original cluster. For the 730-nucleotide mRNA, a full scan without the prefilter took 3 h 43 min. The algorithm time scales as O(mN^3) and memory as O(N^2), where m is the number of candidates and N is the folding size.
What is Tamarind Bio?
Tamarind Bio is a no-code bioinformatics platform built to give life scientists and researchers access to powerful computational tools. Many cutting-edge machine learning models are hard to deploy and use. Tamarind provides an intuitive, web-based environment that removes the complexity of high-performance computing, software dependencies and command-line interfaces.
The platform is designed for biologists, chemists and other researchers who may not have a background in programming or cloud infrastructure but want to run models on their own data. Key features include:
A user-friendly graphical interface for setting up and launching experiments
A robust API for integration into existing research pipelines
An automated system for managing and scaling computational resources
Tamarind treats information and data security as a top priority, as detailed in its Trust Center and Terms of Service.
Accelerating Discovery with OligoWalk on Tamarind Bio
siRNA candidate selection: shortlist siRNAs for a gene of interest ranked by predicted probability of efficient silencing, reducing trial and error.
Accessibility-aware design: use target self-structure and oligonucleotide structure terms to favor accessible binding sites.
Candidate diagnosis: inspect the free energy terms to understand why a sequence is predicted to work or fail.
Targeted region scans: redefine the scan region when only a specific part of the target is of interest.
How to Use OligoWalk on Tamarind Bio
Open the tool: log in to tamarind.bio and select OligoWalk.
Provide the target: enter the target mRNA sequence. The original server accepts A, U, T, G and C, converts T to U, and allows up to 10,000 nucleotides.
Set the oligonucleotide: keep the default 19-nucleotide siRNA, or change length and concentration, if exposed. Concentration does not change the ranking, because the SVM uses standard free energy changes.
Choose the binding mode: the default refolds the target with a partition function calculation. Other modes are faster but less rigorous, if exposed.
Keep the prefilter on: the authors recommend it to save considerable computation time, and you can narrow the scan region if needed.
Run and download: submit the job and retrieve the siRNA efficacy table and thermodynamics table.
Interpret: rank candidates by probability of efficacy and review terms such as Break-targ and End_diff.
Validate: take top candidates to experimental testing or further off-target checks.
Things to Keep in Mind
The positive predictive value of 0.786 means that roughly one in five selected siRNAs may still fail to reach 70% silencing.
The SVM was trained on human cell data at 37 C, so predictions for other conditions may be less reliable.
Computation is expensive: runtime scales as O(mN^3), and the folding size is limited to 1000 nucleotides on the original server.
The paper does not address off-target effects, so these need separate evaluation.
Source: Lu ZJ, Mathews DH. OligoWalk: an online siRNA design tool utilizing hybridization thermodynamics. Nucleic Acids Research 2008, 36 (Web Server issue), W104-W108. DOI: 10.1093/nar/gkn250. Server: http://rna.urmc.rochester.edu/servers/oligowalk.