Oligonucleotides
Working on an siRNA or ASO program?
Our AI models are chemistry-aware: they account for the chemical modifications that drive the potency of real therapeutic oligos, and for the biological context of the target site. We rank your candidates in silico, so you synthesize only the most promising ones.
- siRNAs and antisense oligonucleotides (ASOs), including heavily modified chemistries
- Published with Biogen and Ionis: how phosphorothioate chemistry shapes ASO stability and nuclease resistance, so modification patterns can be chosen before synthesis
- Potency ranking, with a confidence estimate for every prediction
- Models fine-tuned on your own data to learn your chemistry
- Physics-based simulations to guide design when no experimental data exists yet
- Aptamers, microRNAs and long non-coding RNAs
XNAHub® Platform
Modeling natural and non-natural nucleic acids
XNAHub®, our proprietary platform, combines AI and physics-based modeling to design and simulate the sequences and 3D structures of next-generation RNA and DNA therapeutics. It can be deployed in your own environment, so your sequences stay behind your firewall.
- 2D and 3D structure prediction, including chemically modified sequences
- Automatic parameterization of non-natural nucleotides
- Plain and biased MD simulations and free energy calculations
- Built on deep nucleic-acid simulation expertise: Nostrum scientists are part of the Ascona B-DNA Consortium that mapped the physical code of DNA (Nature Communications, 2026)
- Analysis of RNA interactions in protein- or ligand-bound complexes
- Visual workflow builder, molecular viewer and results analysis in one interface
- Bespoke AI models trained on your data
Small Molecules Targeting RNA
Targeting RNA with small molecules?
RNA is far more flexible than a protein, and binding does not always mean function. PELE®, our proprietary Monte Carlo engine, moves the RNA and the ligand together to predict how compounds bind, and molecular dynamics shows whether a binder actually changes the RNA’s functional shape.
- Binding site identification on free and ligand-bound RNA
- Binding pose prediction with induced-fit refinement
- Binding free energy and kinetics (kon/koff)
- Functional profiling: does the ligand lock the RNA in an inactive conformation?
- Optimization of RNA–ligand interactions
mRNA Therapeutics and Vaccines
Developing an mRNA therapeutic or vaccine?
Our AI models design and rank untranslated regions (UTRs) to boost translation and avoid decay, before anything is synthesized. They build on ChaRNABERT, our RNA foundation model, published at ICLR 2025.
- 5′ UTR design for translation efficiency
- 3′ UTR screening for destabilizing elements
- Ranking of new designs against natural human UTRs
- Lipid nanoparticle (LNP) formulation and advanced delivery systems
- Conjugated mRNA platforms
Force Fields
Simulations you can trust
Our team has developed nucleic acid force fields for over 20 years, with more than 3,500 citations. We build on parmbsc2, the state-of-the-art force field for natural and xeno nucleic acids, which combines quantum mechanics and machine learning and has been extensively validated against experimental data.
- Custom parameterization of any new chemistry
- From quantum-mechanics charges and torsions to validated molecular mechanics parameters
- Benchmarking against experimental data, including NMR
- Validation in collaboration with experimental partners
Analysis and Outcomes
From simulation to decision
We combine MD simulations, enhanced sampling and AI to profile and optimize the key properties of your nucleic acid modalities, in a design loop that improves over your program: models rank, the lab tests, and the results retrain the models.
- Thermodynamic stability and folding prediction
- Binding affinity prediction, validated with experimental partners
- RNA–protein, RNA–RNA and DNA–RNA interactions
- Target recognition and specificity
- Sequence–function reverse engineering
- Active learning cycles between in silico design and experiments
