Science
Computation decides which experiments are worth running.
Vivofold works where discovery is most expensive to get wrong: which target to pursue, which molecule to make and which patients will respond. We combine machine learning, physics-based simulation and multi-omics data, and every recommendation comes with the evidence and the uncertainty behind it.
Areas of work
Six areas, one set of methods.
The same data infrastructure, models and simulation tools run through all of them, so an insight in one area can be tested quickly in another.
- 01Small-molecule discovery
Generative AI & physics-based drug discovery
Generative models explore chemical space; physics-based simulation decides which ideas deserve a chemist’s time.
Docking & virtual screening QSAR & cheminformatics Research area
- Structure- and ligand-based design
- Docking and virtual screening
- Molecular dynamics and free-energy methods
- QSAR, ADMET and synthesisability models
- Active learning from assay data
- 02Design from first principles
De novo molecular & protein design
Computational design of new molecules and proteins against a defined target profile, ranked with the same physical models we use in discovery.
- Target-profile driven design workflows
- Computational ranking of candidates
- Design-to-assay feedback with partner laboratories
- 03Systems biology
Multi-omics integration & spatial biology
Genomes, transcriptomes, proteomes and tissue maps analysed together, so a signal is checked from several angles before anyone acts on it.
- Bulk, single-cell and spatial transcriptomics
- Proteomics and metabolomics
- Batch correction and cross-assay integration
- Pathway and network analysis
- Reproducible, documented pipelines
- 04Novel biology
First-in-class target discovery
Evidence-ranked target hypotheses from human genetics, perturbation data and biological networks, with the uncertainty stated.
- Human-genetics and expression evidence
- Network and knowledge-graph prioritisation
- Target dossiers with ranked evidence
- Experimental validation plans
- 05Platform
AI platform for biopharma
One reproducible platform from raw data to decision: pipelines, models, dashboards and reports that a discovery team can rerun and audit.
- Data engineering and FAIR data
- Versioned pipelines and models
- Cloud and high-compute workflows
- Browser-based analysis through Nayancode
- 06Next-generation modalities
Novel biologics: ADCs, mRNA, cell & gene therapy
Computational and data-analysis support for programmes in antibody–drug conjugates, mRNA medicines and cell and gene therapy, delivered with partner laboratories.
- Target and biomarker evaluation
- Patient-stratification analysis
- Developability and CMC data analysis
How a campaign narrows
Most ideas should fail early, and cheaply.
Fast learned models remove most candidates at little cost. Rigorous physics is reserved for the shortlist, where accuracy changes the decision. Assay results then return to the models, so the next round starts from what the last one taught us.
- Ideas generatedGenerative models explore chemical space against the target profile
- Fast filtersDocking, learned affinity and ADMET models remove weak ideas
- Physics rankingMolecular dynamics and free-energy methods on the shortlist
- Assay-readySynthesisable candidates with evidence for each choice
Modalities
One platform across modalities
The same data, simulation and learning infrastructure supports programmes across modalities. Detail for each programme is shared with partners under confidentiality.
Working with us
A pilot first, then a programme
Scope
We agree the target, the product profile, the data you hold and what a good answer looks like.
Pilot
A short, fixed-scope pilot on your data, with the method and the uncertainty written down.
Deliver
Ranked results, code, models and a methods section your team can rerun and defend.
Iterate
Assay results go back into the models, and the next cycle starts from what was learned.
Also from Vivofold
The people and the software behind the science
Most engagements begin with a short, fixed-scope pilot on the partner’s own question and data. If the pilot is useful, it becomes a longer programme with agreed milestones. Methods, code and models are documented so that your team can rerun and audit them.
Yes, under a confidentiality agreement. Partner data, and models trained on it, are used only for that partner’s programme.
Vedicgen is our school and Nayancode is our scientific software. They are part of the same company: Vedicgen trains the people who do this work, and Nayancode is the software we build for it, including free tools you can use today.
Yes. The Nayancode Workbench and free tools are open to anyone, and Vedicgen courses and internships take applications from students and professionals.
Partner with Vivofold
Bring us a target, a dataset or a question.
Tell us what you need to decide. We reply with a proposed pilot, a timeline and a fixed price.