For drug developers, a protein’s three-dimensional shape is everything. It dictates how a potential therapeutic will bind to its target, or whether it will misfold into a useless,or toxic,clump. The dominant tools for predicting that structure are now massive AI models, trained on millions of known protein sequences and their solved structures. But what happens when you need to design a protein that has never existed in nature? A university spinout in the Czech Republic is making a quiet, foundational bet that the answer lies not in more data, but in a deeper code.
MendelFOLD, incorporated in late 2024, is developing computational methods to predict the 3D structure and function of what it calls “biorelevant” proteins. Its scientific premise is a concept its founders have worked on for decades: a universal proteomic code. This is not a large language model for proteins. It is a proposed set of physical and chemical rules, derived from the interactions between amino acid residues, that the company believes can fundamentally explain and predict how a linear chain of amino acids folds into a functional, three-dimensional machine [PERPLEXITY SONAR PRO BRIEF]. For a field increasingly reliant on pattern recognition in existing data, it is a deliberate turn back toward first principles.
The bet on a fundamental rulebook
The company’s core intellectual property stems from long-standing academic work by its co-founder and chief scientific officer, Andrew David Miller. A professor of organic chemistry and chemical biology, Miller has published on the concept of a proteomic code since at least 2002, exploring how information for protein folding might be embedded within nucleic acids and the interactions between sense and antisense peptides [The proteomic code - Brno, May 2025]. Recent peer-reviewed papers, co-authored by the MendelFOLD team, describe novel amino acid residue pairing models that they argue “encode” protein folding and protein-protein interactions [The proteomic code: Novel amino acid residue pairing models "encode" protein folding and protein-protein interactions - PubMed, 2026].
The ambition is to build a predictive platform that combines this theoretical framework with molecular simulation. The stated wedge is a knowledge-based approach intended to address limitations of purely data-driven AI methods, particularly in designing novel proteins for synthetic biology or therapeutics where no natural analogue exists [PERPLEXITY SONAR PRO BRIEF]. If the code holds, it could provide a more interpretable, physics-grounded roadmap for protein engineering.
An academic engine in Brno
MendelFOLD is a classic deep-tech spinout, its roots deeply embedded in Mendel University in Brno. The founding team reads like a department roster, blending decades of academic leadership with newer computational expertise.
| Role | Name | Primary Affiliation & Background |
|---|---|---|
| Co-founder, CSO | Andrew David Miller | Professor of Organic Chemistry & Chemical Biology, Mendel University; former professor at Imperial College London [Andrew David Miller, Professor, Chief Scientific Officer - eMedEvents, 2026]. |
| Co-founder | Zbyněk Heger | Head of the Department of Chemistry and Biochemistry, Mendel University; research focuses on cancer, drug discovery, and nanomedicine [Zbynek HEGER |
| Co-founder, Head of Bioinformatics | Tareq Yakoub Hassan Hameduh | Postdoctoral researcher in computational protein science at Mendel University [PERPLEXITY SONAR PRO BRIEF]. |
| Director | Andrew Hladký | Listed company director [PERPLEXITY SONAR PRO BRIEF]. |
| Director | Nicholas Geoffrey Alan Weaver | Listed company director [PERPLEXITY SONAR PRO BRIEF]. |
The company’s formation was facilitated by IP Lab Ventures, a Czech venture-building and technology-transfer firm focused on deep-tech in Central and Eastern Europe. IP Lab lists MendelFOLD among the spin-offs it helped establish, though the precise nature of its financial backing is not publicly detailed [IP Lab Ventures]. The company also received non-equity support from EIT Health through the InnovPrecMed initiative, which nominated it for a Falling Walls Venture showcase in 2026 [EIT Higher Education Initiative / Falling Walls Ventures, May 2026].
The long road from code to clinic
The vision is grand, but the path from a theoretical code to a validated drug discovery tool is measured in years and clinical phases. The company has indicated its initial target customers are pharmaceutical and synthetic-biology companies, with plans to pursue research collaborations [PERPLEXITY SONAR PRO BRIEF]. No named commercial partnerships or deployments have been announced. The primary traction signals so far are academic: continued publication in peer-reviewed journals and recognition from European deep-tech and health innovation networks.
The competitive and technical risks here are substantial. The field of computational protein design is crowded with well-funded AI-native players whose models have delivered staggering results on public benchmarks. MendelFOLD’s approach, by its nature, cannot move as fast as a model that ingests the entire Protein Data Bank. Its success hinges on the proteomic code being both fundamentally correct and computationally tractable enough to outperform or usefully complement data-driven methods.
- Proof-of-concept gap. The published science establishes a hypothesis and early models, but not yet a validated platform that can reliably design a de novo therapeutic protein. Bridging that gap requires significant further software development and rigorous, blinded testing against industry standards.
- Commercial runway. As a pre-seed company with undisclosed funding, its resources for the multi-year R&D effort ahead are unclear. Its ability to attract the computational biology talent needed to build a robust product will be tested.
- Regulatory context. Any tool intended for therapeutic protein design will eventually need to demonstrate its predictive accuracy under the scrutiny of regulators like the FDA or EMA, a bar that purely research-focused models never face.
The company’s most plausible answer is that it is not trying to beat AI at its own game, but to play a different one. For certain protein engineering puzzles,especially those involving non-natural amino acids or entirely novel folds,a first-principles understanding could be the only viable starting point.
What to watch in the next 12 months
For a company at this stage, milestones are less about revenue and more about validation. The next year will likely focus on translating its published research into a more concrete software prototype and securing its first industry partnership. A collaboration with a pharmaceutical or synthetic biology firm, even a non-exclusive research agreement, would be a critical signal that the proteomic code has practical utility beyond academia. Given its academic origins and support from IP Lab Ventures, a larger seed round to fund this translational push seems a probable next step.
The ultimate patient population for this technology is broad, encompassing anyone waiting for a new biologic therapy or a more efficient enzyme for industrial manufacturing. Today, the standard of care for protein-based drug discovery remains a slow, expensive cycle of hypothesis, wet-lab synthesis, and experimental structure solving. Computational tools have accelerated parts of this process, but the dream of reliably designing a protein from scratch for a specific function remains largely unrealized. MendelFOLD is betting that the key to unlocking that dream was written in biology’s fundamental rules all along, waiting to be decoded.
Sources
- [The proteomic code - Brno, May 2025] The proteomic code | https://www.proteomic-code.com/
- [The proteomic code: Novel amino acid residue pairing models "encode" protein folding and protein-protein interactions - PubMed, 2026] The proteomic code: Novel amino acid residue pairing models | https://pubmed.ncbi.nlm.nih.gov/
- [Andrew David Miller, Professor, Chief Scientific Officer - eMedEvents, 2026] Andrew David Miller profile | https://www.emedevents.com/
- [Zbynek HEGER | Head of Department | Assoc. Prof. Ph.D. | Mendel University in Brno, Brno | Research profile, 2026] Zbyněk Heger research profile | https://www.researchgate.net/
- [IP Lab Ventures] IP Lab Ventures portfolio | https://www.iplventures.com/
- [EIT Higher Education Initiative / Falling Walls Ventures, May 2026] MendelFOLD - What if we could understand life before it breaks? | https://www.linkedin.com/posts/eit-higher-education-initiative_falling-walls-venturescoders-that-cure-activity-7458070785019039744-k0mP