This is how we engineer better proteins

Applications

Antibody Design

Enzyme Engineering

Vaccine Design

From antibodies and vaccines to enzymes and de novo binders, we design and engineer proteins that perform. On every project we benchmark open-source and proprietary protein language models and AI algorithms, then validate the best designs with physics-based modeling, so you test fewer sequences and reach better candidates faster.

Antibody Discovery

Designing an antibody from scratch?

We cover the full pipeline, from target discovery to de novo design, de-risking and ranking, across IgG, Fab, scFv, VHH, VNAR and bispecific formats. Our clients typically reach a promising candidate by testing fewer than 40 sequences in the wet lab.

  • Target discovery with Yggdrasil, our knowledge graph linking 2.47 million diseases, targets and compounds
  • De novo design with diffusion models, antibody language models and inverse folding, including our own NostrumAbMPNN, benchmarked against state-of-the-art open-source tools
  • De-risking with in-house predictors of aggregation, CDR flexibility, immunogenicity and T-cell epitopes (PredIG)
  • Validation and ranking with consensus AI co-folding, PELE® with Loop Closure and molecular dynamics free energies
  • Affinity maturation of existing antibodies and epitope characterization

Vaccine Design

Vaccines designed for sequential and conformational epitopes

Starting from a known neutralizing antibody, we design immunogens that present its epitope in its native conformation. By integrating AI, structural biology and high-performance computing, we bring the speed and precision needed to stay ahead of global health threats.

  • VRSVAC with HIPRA, which delivered successful vaccine candidates
  • EpiVac, focused on epitope discovery for Porcine Reproductive and Respiratory Syndrome Virus (PRRSV)
  • RAPID-VAC, an automated platform to design a vaccine candidate against any “Pathogen X” in just one week
  • Epitope mapping: hot spots, mechanism of action and competing epitopes

De novo design of binding proteins

Need a binder for a difficult target?

We design small and medium-sized protein binders de novo for any protein of interest, from GPCRs to circulating proteins. Generative AI proposes backbones and sequences, and physics-based modeling ranks them, giving high-quality designs at a lower computational cost than conventional methods.

  • Affilin® binders for HER3, designed with Navigo Proteins GmbH: of 12 sequences tested, 7 expressed and bound, 2 below 100 nM, in 1–2 months instead of 6 months by phage display (International Journal of Molecular Sciences, 2025)
  • International Nipah virus glycoprotein binder competition: 2nd of 680 participants in in silico design and 10th in de novo design, and our binder showed the best expression
  • Open-source and proprietary models benchmarked on every target, so each design comes from the best-performing method

Enzyme Bioprospecting

Looking for an enzyme that performs your reaction under specific conditions?

Our AI and physics-based pipelines screen thousands of sequences in silico to find the right enzyme quickly.

  • Our proprietary protein language model, fine-tuned to your enzyme family to generate new sequences
  • Smart mutation libraries to focus your experimental effort
  • Multifunctional enzymes with de novo active sites that support multiple chemistries (PluriZymes), published in Nature Catalysis
  • New functions for existing proteins and production strains, without foreign DNA: a native E. coli protein redesigned to break down PET nanoplastics (Trends in Biotechnology, 2025)

Computational protocol to design a hydrolase site

The video shows how we use PELE to explore non-functional binding sites in proteins and functionalize them through site-directed mutagenesis by adding the catalytic triad of a serine hydrolase.

Sergi RodaDirector of Protein Engineering

Have a project in mind? Get in touch and I’ll be happy to discuss how we can help.