Innovate biology
Minibinder design and explainable bioinformatics for hard protein targets. Start with a written target assessment; we design candidates and coordinate partner-laboratory testing.
Accessibility, immune tolerance, specificity and antigen quality create different obstacles to binder discovery. Each example shows a question we assess before choosing a design and testing strategy. Our work includes membrane and soluble proteins.
A, accessibility-limited
A small, recessed or partly covered epitope can limit which binder shapes reach it. We assess accessibility in the relevant protein structure and biological context.
Shown: a multi-pass membrane protein, where only the short extracellular loops are accessible in the native fold.
B, immune tolerance-limited
A protein conserved between human and mouse may elicit little immune response. Computational design starts from its structure, without relying on an animal to produce antibodies.
Shown: a protein conserved to the residue, since human and mouse calmodulin share one sequence.
C, discrimination-limited
Closely related proteins can share the same binding surface. We include the proteins to avoid in the assessment and computational comparison; selectivity is established by the agreed experimental tests.
Shown: two paralogues that differ at a handful of positions, β-actin and γ-actin.
D, antigen supply-limited
Misfolding, aggregation or poor expression can compromise a binding assay. A structure or model supports an initial assessment, while representative target material remains essential for experimental testing.
Shown: a flexible stretch held in the membrane, from the amyloid precursor protein.
We support your complete biological data journey: from omics analysis and explainable modelling to target assessment, protein design, experimental testing and scientific applications. These services form a closed loop, with new data and results informing the next round of work. You receive the evidence behind each decision and can commission the complete cycle or individual services.
By us, in silicoThrough a partner laboratory
Before: find the target
Our focus: the binder for a hard target
After: read the result
Binder tests and proteomics feed the next round
From small data to big data, with one contract and one contact. Order one service, or several in a row. We select the laboratory for your assay, coordinate the work and report the result. We are still contracting our partners and name them per project.
AAanalysis, our open framework, connects sequence predictions to physicochemical properties and residue positions. These features support a selection you can inspect and hypotheses you can test.
Published and open. γ-Secretase substrates, AAontology, AAclust, and the AAanalysis package.
The published methods make the basis of a prediction available for review and experimental testing.
Published result
Predicting which proteins γ-secretase cuts, our published Comparative Physicochemical Profiling (CPP) framework beat embeddings from a state-of-the-art protein language model. Every feature it uses is a named property at a named position.
60%→90%
Prediction accuracy, against a protein language model
88%
15 of 17 predictions confirmed experimentally
Breimann et al., Nature Communications 16, 5428 (2025). 10.1038/s41467-025-60638-z. The methods are our own and open: AAontology, AAclust, AAanalysis The published result concerns γ-secretase substrate prediction. Read the study for its benchmark design.
Academic groups and biotech companies, starting in the German-speaking region.
A defined target and limited capacity for a design campaign. Commission the assessment first, then decide whether to proceed with candidate design and testing.
A research reagent, or an answer before you commit. Scope and rights are settled before the work starts.
A biological question and data ready to analyse. Receive interpretable results, figures and a pipeline your group can run again.
You get a straight answer about what is realistic.
Research use only. Scope, criteria and deliverables are agreed per project.
Built at the Technical University of Munich
Where our team studied and our advisors work
