An enzyme that works at your conditions, not in a model organism.
Innophore designs and engineers enzymes for industrial biotech across flavor and fragrance, consumer goods, food, biomaterials, biosensors, biofuels, environmental and diagnostics. Cavity-first computational discovery, structure-guided engineering for activity, selectivity and stability, and partner wet-lab follow-up when needed. Customer keeps the IP on every enzyme delivered.
What we can do, and how often it works.
A decade of cavity-first enzyme discovery and engineering, with a hit rate that scales with how much you already know about your target.
Of the enzyme world, covered.
Five of the six classical EC classes: oxidoreductases, transferases, hydrolases, lyases and ligases, plus specialty work on polymerases. Every major enzyme class in industrial biocatalysis, with cavity-first methods.
Oxidoreductases
- Oxygen-introducing reactions (peroxidases, oxygenases, CYP P450s)
- Cofactor-dependent reductions (KREDs, IREDs, RedAms, ene-reductases)
- Flavin-dependent oxidoreductions (flavoproteins)
Transferases
- Amine-group transfer (transaminases)
- Methyl-group transfer (methyltransferases)
- Glycosyl-group transfer (glycosyltransferases)
- Phosphate-group transfer (kinases, phosphotransferases)
- Nucleoside transfer (nucleoside transferases)
Hydrolases
- Proteases, lipases, esterases
- Nitrilases, amylases
- Glycoside hydrolases
- Phosphatases
Lyases
- Water-removing & -adding enzymes ((de)hydratases)
- Cyclases & terpene lyases
- Decarboxylases
- Aldolases
Ligases
- DNA & RNA ligases
- ATP-dependent ligases
- Peptide ligases
Polymerases
- DNA & RNA polymerases
- Engineered variants for non-natural substrates
- Reduced byproduct formation
How often we find what you are looking for.
The hit rate scales with how much prior information is available about the target reaction. The more you know going in, the higher the chance of a productive hit.
From a reaction sketch to an enzyme your process can use.
Industrial biocatalysis only pays off when the enzyme survives the reactor. Most projects pair a discovery phase with engineering for the property that actually limits the process, whether that is thermostability, pH range, selectivity, or activity on a non-native substrate.
Discovery
Cavity-based search across PDB, customer sequence space, internal homology models and public metagenomes (MGnify and equivalents). Ranked enzyme candidates come with a structural rationale, beyond what a sequence-only BLAST would return.
Engineering
Focused variant libraries for activity, substrate scope, selectivity, thermostability or pH range. Built with in-silico mutagenesis, MD validation, ancestral sequence reconstruction and generative redesign tools where they help.
Genome and proteome work
Annotation of customer genomes and proteomes by cavity function. SeqScan of all ORFs in six reading frames, full structural modeling and cavity-based functional matching against query catalytic motifs.
Every project leaves the customer with a dedicated Catalophore™ instance, the structural files behind every recommendation, and the IP on every enzyme and variant we deliver.
How we run industrial biotech engagementsDefined intake. Structural work. Hand-off to your bench.
A typical engagement is one to three work packages. We start from a target reaction or a customer-supplied sequence and finish with files that go straight into expression, screening or formulation work.
Brief and intake
Customer provides the target reaction, substrate, internal sequences or genome data, plus the process constraints (pH, temperature, solvent, scale). Confidential information is scoped from the start.
Catalophore™ work
Sequence space analysis, structural modeling, 3D point-cloud cavity matching against PDB and curated databases, MD validation, in-silico mutagenesis and ML-assisted ranking. Reviewed by senior scientists.
Hand-off and wet-lab
Ranked candidates with PDB and PyMOL files, mutation lists and a written report. Optional production and testing of variants through partner CMOs in E. coli, Pichia and bioreactor scale up to 3 L pilot.
Industrial biocatalysis fails for measurable reasons. An enzyme that works at 25°C and pH 7 is a different molecule from one that needs to run at 40°C, pH 5 and 200 mM substrate. AI structure prediction is useful but still gets the chemistry wrong often enough to matter, with stereochemistry errors in 4.4% of cases by one benchmark.
Every structure that informs a Catalophore™ recommendation is energy minimized, refined and reviewed before it goes into a customer report. If the model is wrong, the enzyme list is wrong.
Anonymized examples from recent industrial-biotech projects.
Nine representative engagements across flavor and fragrance, consumer goods, food, biomaterials, biosensors, biofuels and environmental. Customer names and target molecules are generalized. Project scope and deliverables are reported as run.
A four-year enzyme discovery program
A global flavor and fragrance house needed to expand its proprietary enzyme library for selective biosynthesis of ingredients. We built a continuous service running sequence-space expansion of in-house enzymes, homology modeling, Catalophore™ cavity and point-cloud computation, MD simulations and substrate docking. The relationship has been renewed every year since.
Enzyme system for a cleaning formulation
A consumer brand needed enzymes able to dissolve stubborn organic residues by combined reductive and proteolytic action. We ran a Catalophore™ in-silico search for specialty protein-degrading enzymes (BLAST plus modeling of around 2,000 representatives plus point-cloud matching), and followed up with an in-silico engineering work package using MD-guided mutagenesis and a focused variant library. A scale-up package added Pichia fed-batch production in 3 L bioreactors with downstream concentration and lyophilization.
Glycoside hydrolase redesign for laundry
A global laundry and home-care leader needed an engineered enzyme variant that kept activity while moving down the sequence-identity axis to enable freedom-to-operate. We ran cavity and surface Halo analysis, generative ML protein redesign coupled with cavity-AI models, and stability and expressibility scoring.
Three enzymes for continuous monitoring
A continuous-biosensor manufacturer needed three different enzyme systems over four years, each for a different small-molecule biomarker. Each project combined sequence-space analysis, cavity comparison across related enzyme families, proteome mining of 4,000 to 5,200 candidate microbial sequences, and MD-supported mutagenesis.
Mycotoxin detoxification under defined process conditions
An agricultural processor needed enzymatic detoxification of two distinct mycotoxin classes under defined fermentation conditions (mildly acidic pH and moderate temperature, aerobic). We ran cavity-similarity screening over PDB, the in-house cavity database and MGnify metagenomes, ranking by cavity volume, electrostatics and environmental tolerance metadata.
Three-enzyme cascade for an oil refinery
A grain and oilseed processor needed to suppress contaminant formation during high-temperature edible-oil deodorization. We ran two parallel searches. The first looked for low-water-activity lipases that lower the contaminant precursor by esterification. The second built a three-step enzyme cascade from complementary hydrolase and dehalogenase families with tunnel engineering and docking at the oil-water interface.
Silk fiber properties improved
An industrial biomaterials manufacturer needed predictable control of fiber properties in protein-based silk fibers for industrial applications. We built a custom ML pipeline combining evolutionary polynomial regression, a domain-specific protein language model and physics-based nanostructure models, trained on public sequence data.
Genome mining for second-generation feedstocks
A bioethanol producer wanted to convert sugars and improve fiber breakdown from its internal microbial genome data. We translated and curated the genomic data, ran targeted mining for sugar-converting and glycoside hydrolase enzymes, modeled active sites with Catalophore™ cavities and matched them against query catalytic motifs. A second project searched for proteases and accessory hemicellulases for corn-oil recovery from wet cake.
Metal-binding proteins
A biomining and critical-metals company needed proteins with improved selectivity for a target metal ion over background metal binding. We ran cavity and Halo surface analysis on literature metal-binding proteins, MD on coordinating residues and deep-learning segmentation of ion-binding sites trained on Catalophore™ cavity data, with an optional discovery work package across NCBI and MGnify.
Patented method, ten years of biocatalysis publications.
The cavity point-cloud method that drives every project is a granted patent. The same method has discovered new biocatalysts in peer-reviewed work since 2014, the year of the founding Catalophore™ publication.
Protected core method
The point-cloud cavity method is a granted patent (US 2015/0302142 A1, also granted as US 10,825,547 B2) and protected across 15 jurisdictions.
100x on production GPUs
Cavity matching throughput accelerated by a factor exceeding 100 on enterprise GPU systems. A proteome match that took roughly 625 seconds now completes in about 5.
Official NVIDIA partner
Listed NVIDIA healthcare and life sciences partner. Public reference deployments built on BioNeMo (proteome dataset built in two weeks against a year-long baseline).
Cavity-based enzyme discovery
Peer-reviewed proof that cavity matching finds new biocatalysts. The founding Catalophore™ paper (Nature Communications, 2014) and recent work on fatty acid photodecarboxylases (ChemBioChem, 2024) demonstrate the method generalizes across protein space.
The same cavity search that discovers new biocatalysts for fragrance or food is the one that finds new fatty acid photodecarboxylases in academic work. One method, validated across protein space.
Cavity-based biocatalyst discovery, published
One platform, four ways to work with us.
Industrial biotech customers usually start with a single work package, then either grow into a multi-year service contract, run a one-off engineering campaign or take the platform in-house. Every engagement includes a customer Catalophore™ instance and full IP transfer on enzymes and variants.
- WP
Work package projects
Fixed scope, defined deliverables, six to twelve weeks. Most common entry point for new partners.
- FTE
Annual service contracts
As many FTEs as the customer needs, allocated to a continuous program. Renewed yearly. Common with customers who run repeat enzyme campaigns each season.
- IP
IP on enzymes
Intellectual property on identified, built or modified enzymes belongs to the customer in every engagement model.
Typical engagement shapes
- Pilot Single work package One enzyme family, one substrate or one process constraint. Six to twelve weeks. Ranked candidate list with structural rationale on a dedicated Catalophore™ instance.
- Program Multi-WP campaign Discovery work package followed by engineering rounds, optionally with wet-lab follow-up via partner CMOs from shake-flask through 3 L bioreactor. Monthly review.
- Annual service Customer-defined number of FTEs Continuous-collaboration model with a customer-branded Catalophore™ instance. The customer decides how many FTEs to allocate. Monthly fee. Suitable for repeat-need customers in fragrance, food and consumer goods.
- On-premise Standalone deployment Catalophore Pro on a dedicated AI workstation preloaded with the PDB cavity database and a curated customer enzyme dataset, plus a remote training workshop.
Bring us a reaction.
Bench scientists, process engineers and IP teams come in through different doors. Both lead to the same scientists.
Start a project
Send us a reaction, a substrate, a sequence or a genome. We come back with a scoped pilot proposal and an honest call on feasibility.
Send a brief →Explore on-premise
Take a curated Catalophore™ deployment in-house on dedicated hardware, with the public reference database and your own enzyme dataset preloaded.
Talk deployment →