AI-Native RNA Design Platform
Design therapeutically active RNA at molecular precision
From sequence to 3D structure prediction, energy-based ranking and sequence optimisation — all in one AI-driven platform built for RNA drug discovery. EU based.
// Live 3D
From sequence to a real 3D fold
This is an actual model prediction for an RNA-Puzzles blind target — near-native fold at experimental precision, rendered straight from the reconstructed backbone.
// Core capabilities
Three models, one unified pipeline
Our platform integrates structure prediction, thermodynamic ranking and inverse design into a single coherent workflow for RNA therapeutic discovery.
// How it works
From sequence to ranked candidate in minutes
Our three-stage pipeline replaces weeks of manual iterative design with a computational workflow that runs in your browser or via API.
Input your sequence
Paste or upload any RNA sequence. Define optional constraints: GC content range, modification sites, target binding regions.
AI predicts and ranks
Our neural network predicts full 3D tertiary structure. The energy model scores thermodynamic stability. The inverse design module generates optimised sequence variants.
Export ranked candidates
Receive a ranked library of optimised sequences with 3D coordinate files, energy scores and confidence metrics, ready for wet lab validation.
// Why RNAfold
Built for RNA, not adapted from protein tools
Most computational biology platforms were designed for proteins and retrofitted for RNA. We built from the ground up for the unique properties of RNA therapeutics.
Native thermodynamic modelling
Energy-based ranking uses RNA-specific free energy parameters, not generic molecular mechanics. More accurate stability prediction for therapeutic RNA modalities.
Tertiary structure, not just secondary
Go beyond dot-bracket notation. Our model predicts full 3D coordinates, essential for drug binding site analysis and delivery optimisation.
Inverse design built in
Design backwards from a structural target. Gradient-based optimisation explores sequence space orders of magnitude faster than screening.
Pipeline-native integration
REST API with standard bioinformatics output formats. Fits into your existing LIMS, ELN or custom discovery pipeline.
EU-based, GDPR-compliant
Incorporated in Malta. All data processing within EU jurisdiction. No sequence data leaves European infrastructure.
Continuously improving model
Our model improves with every validated sequence in the platform. Customers benefit from accuracy gains as the experimental data pool grows.
// Benchmark
Validated on blind prediction targets
We evaluate on the RNA-Puzzles community benchmark — experimentally determined structures the model never sees during training. Results shown are independent core-fold assessments using PyMOL-equivalent alignment.
RNA-Puzzles P04
Small hairpin
RNA-Puzzles P06
Hairpin / loop
RNA-Puzzles P13
tRNA-sized
RNA-Puzzles P16
Riboswitch core
RNA-Puzzles P05
Complex multi-loop
RNA-Puzzles P15
Pseudoknot
RNA-Puzzles P14
Junction
"Near-native" indicates a fold recovery at experimental precision on the core topology (PyMOL-equivalent iterative alignment, C4′ backbone). Outlier loops excluded per standard practice. Full per-target breakdown is provided under NDA during pilot evaluation.
// Live demo
See the reconstruction in action
Select a benchmark RNA target and explore the predicted 3D structure.
HIV-1 TAR RNA
Conformational flexibility benchmark. Two binding-competent states relevant to Tat protein inhibition and antiviral drug design.
FMN Riboswitch
Bacterial gene-expression regulator with characterised apo and holo conformations. Target of antibiotic Ribocil.
RNA-Puzzles P16
Blind prediction benchmark. Demonstrates near-native fold recovery at experimental precision.
// Get started
Ready to accelerate your RNA design workflow?
We are onboarding our first pilot customers now. Request access and we will schedule a 30-minute demonstration on your sequences.
Early access · Pilot pricing · Malta-based · EU data residency