Catalophore
Catalophore is Innophore's protein cavity software suite. It maps every binding site as a 3D physico-chemical point cloud, so research teams can search whole proteomes by what a pocket does rather than by sequence or fold.
What is Catalophore?
Catalophore is the software family of Innophore GmbH, an Austrian company spun out of the acib research centre and the University of Graz. Every product runs on the same patented idea: a protein cavity is described as a three-dimensional point cloud, each grid point carrying physico-chemical values such as hydrophobicity, electrostatics, desolvation, hydrogen-bond character or metal coordination. Because the comparison works on the shape and chemistry of the pocket, two proteins with no evolutionary relationship can still be recognised as functional neighbours, which sequence alignment and fold comparison cannot show.
The range has four members. Catalophore Platform Pro is the full workbench installed on the customer's own site: 10.2 million binding sites indexed, 231,000 PDB structures parsed and 574,000 Swiss-Prot entries mirrored locally, with web interface, job scheduler, gateway, file server, licence authority, workers and database all running behind the firewall, air-gapped deployments included. Catalophore Studio is a browser-based cavity workbench that applies the Caviar detection algorithm, annotates more than thirty property fields per pocket, returns a structured nine-section AI reading of the site, prepares ligands with RDKit and GAFF2 typing, docks them and searches the cavitome through eighty-dimensional embeddings compared by cosine similarity. CavitOmiX Copilot is the cloud module for off-target work, built on 42,042 human proteins modelled with NVIDIA BioNeMo and offered in a free Basic version and a paid Pro version. Finally, a free plugin brings Catalophore cavity detection and CavitOmiX matching straight into PyMOL.
Scoring is deliberately explainable: the similarity formulas are published, matches carry a Gumbel e-value in the manner of BLAST, property weights are adjustable, and every run is stamped with a UUID recorded in an audit log. A conversational assistant called Cavi can build a project from a DOI, run sandboxed Python and write reports, using a large language model of the customer's choice that is switched off by default. The underlying method was first published in Nature Communications in 2014. This entry covers the software only; Innophore also sells discovery work as a service, which is a separate offering.
What it does
- Detect every cavity in a protein structure as a point cloud annotated with 15 to 30 physico-chemical property fields
- Search millions of indexed binding sites for pockets that match a query cavity, independently of sequence or fold
- Rank matches with a documented similarity formula and a BLAST-style Gumbel e-value
- Predict a protein structure with AlphaFold, ESMFold, OpenFold or classical homology modelling
- Dock a ligand with AutoDock Vina, or generate poses with the diffusion model DiffDock
- Mutate residues in silico across chains, with combinatorial expansion and energy minimisation
- Export a branded PDF, DOCX, XLSX or PPTX report carrying a UUID watermark and a diff since the previous run
When to use Catalophore / When not to
A quick filter to help you decide if Catalophore is the right fit.
When to use Catalophore
- Drug discovery and drug safety teams looking for off-target liabilities before a candidate reaches the clinic
- Protein engineers and biocatalysis groups hunting for enzymes with a specific stability, pH range or substrate scope
- Structural bioinformaticians who want cavity detection, docking and proteome-wide search in one workbench
- Pharma and biotech organisations that require an on-premise or air-gapped deployment behind their own firewall
- IP and regulatory teams that need a scored result they can defend, with an audit trail and a documented formula
When not to use Catalophore
- Individuals or hobbyists: this is a business-to-business tool and its terms are written for corporate customers only
- Buyers who need a published price list, since the flagship products are quoted case by case
- Teams working outside proteins and small molecules, as nothing here covers non-biological chemistry or clinical imaging
- Anyone expecting a self-service trial of the full platform, because Platform Pro and Studio start with a booked demonstration
- Users without structural biology training, who will struggle to read property fields and tune matching weights
How to use Catalophore
A typical end-to-end flow, from setup to results.
- Decide which product fits: cloud copilot, browser workbench, on-premise platform or PyMOL plugin
- For a first free look, create an account on the CavitOmiX Copilot and run the guided example search, which costs no tokens
- Alternatively, download the free PyMOL plugin and install it following the PyMOL Wiki plugin procedure
- Load a structure by uploading a file or fetching it from the PDB, AlphaFold DB or UniProt
- Run cavity detection, then read the property point cloud and the structured AI interpretation of the site
- Point a cavity matching or proteome-wide search at a saved collection to obtain a ranked list of similar pockets
- Prepare a ligand, dock it and review the scored poses and interaction breakdown
- Buy tokens when the free allowance runs out, since each search result, report or structure download is charged
- Export the run as a branded report or as raw PDB, CIF, FASTA or MOL2 files
- For Platform Pro or Studio, request a demonstration or an integration discussion, then deploy the container stack and connect HPC, Docker and GPU nodes
Pros & Cons
Pros
- The underlying method has been peer-reviewed since 2014, with later papers co-authored by NVIDIA, AWS and Google scientists
- Scoring is explainable rather than opaque: published formulas, Gumbel e-values, adjustable weights and a UUID on every run
- Full on-premise deployment is supported, including air-gapped installations with reference data pre-loaded
- Intellectual property in uploaded data and in the results is left to the customer in explicit terms
- Open exports in PDB, CIF, FASTA and MOL2 keep the data usable outside the platform
- A documented REST API and an MCP server allow the tool to be dropped into existing Snakemake, Nextflow, KNIME or Python pipelines
- There is a genuine free entry point: the PyMOL plugin and the Basic version of the cloud copilot
Cons
- No price is published for Catalophore Platform Pro, Catalophore Studio or the Pro version of the copilot: everything goes through a quotation
- The token price grid is only visible after creating an account, since the purchase page is behind authentication
- No data processing agreement, no sub-processor list and no hosting country are published for the cloud module
- No documented way to exclude customer data from model training
- The general terms date from 2018, are primarily German, and describe a classic software supply contract rather than an online service
- Several published links are dead, including the terms PDF, the PyMOL plugin archive and the Google Play listing of the CavAR app
- The interface and documentation are English only, and tokens are non-refundable once bought
Pricing & Plans
There is a permanent free offer: the Catalophore PyMOL plugin is free of charge, and the Basic version of the CavitOmiX Copilot can be used without payment, with a guided example search that consumes no tokens. Paid use of the copilot runs on prepaid tokens, sold in three packs at net prices excluding VAT: 100 tokens for 99 EUR, 500 tokens for 449 EUR and 1,000 tokens for 799 EUR, which corresponds to an entry rate of 0.99 EUR per token falling to 0.799 EUR at the largest pack. Tokens are consumed per action: one token per off-target search result with a minimum of ten results, one token for a report or a single structure download, and ten tokens to download every matching structure at once. Purchased tokens do not expire but are not refundable. Payment is taken by card, GooglePay or Link. No price is published for Catalophore Platform Pro, Catalophore Studio or the Pro version of the copilot, which are quoted on request, and the general terms simply state that all prices are in euros excluding value added tax.
- free
- with binding-site analysis and AI-guided off-target search across the human proteome
- paid actions charged in tokens
- custom pricing
- adding structure prediction and protein design
- proprietary Catalophore calculation
- cross-organism matching
- proprietary ligand upload
- docking and molecular dynamics
- advanced 3D visualisation and priority access to new features
- Token pack of 100 tokens for 99 EUR
- net of VAT
- Token pack of 500 tokens for 449 EUR
- net of VAT
- Token pack of 1
- 000 tokens for 799 EUR
- net of VAT
- single-team pilot on one Facility
- quoted on request
- discovery group with additional seats
- shared collections
- multi-team access control
- extra HPC and GPU Facilities and customer-branded reports
- quoted on request
- organisation-wide rollout with per-facility branding
- federated access control
- automated backups
- archival lifecycle and full audit reporting
- quoted on request
- free plugin
- with an optional contribution to the development fund
Data, GDPR & hosting
A consolidated view of how Catalophore handles your data.
GDPR overview
Innophore is established in Austria, inside the European Union, so no Article 27 representative is required and none is named. Its data protection declaration, dated 1 October 2024, is published in German and English and invokes the EU General Data Protection Regulation. It lists rights of access, erasure, rectification, blocking and objection at any time without justification, and gives customer@innophore.com as the single channel to exercise them. The CavitOmiX Copilot help pages add an explicit claim of compliance with Regulation (EU) 2016/679 and describe SSL/TLS encryption in transit and at rest. Three things are missing: no data protection officer is named, no data processing agreement is published or offered, and no list of sub-processors exists anywhere on the site.
Who owns the data?
Innophore states that the intellectual property of both the input data and the analysis results stays with the user, and that it claims no ownership over research data or analysis outputs. Uploaded data is described as processed only to deliver the requested service. The proprietary algorithms, methods and tools remain Innophore's own property. The Austrian general terms add that, once the fee is paid, the customer receives a non-exclusive, non-transferable, non-sublicensable and unlimited right to use the software on the contracted hardware, while every other right stays with the vendor. For the on-premise platform the whole stack runs on the customer's own site, so the data never leaves that environment.
Reuse rights
Results and uploaded material belong to the customer, who may reuse them freely: exports are offered in PDB, CIF, FASTA and MOL2 alongside branded PDF, DOCX, XLSX and PPTX reports, and Innophore presents this openness as an absence of vendor lock-in. The reverse direction is restricted: users may not scrape, crawl or compile datasets from the Cavitomix platform for any purpose, including academic research and machine learning training, without prior written permission. On the vendor side, the privacy declaration says personal data is never sold and is used to run and improve the service, provide support and perform internal analysis. Web server logs record IP address, timestamp, request and browser data. Anyone using the free PyMOL plugin should note that the input structure or sequence is forwarded to third parties, namely NVIDIA BioNeMo, Meta ESMFold, DeepMind AlphaFold and the RCSB PDB, whose own practices apply.
Data retention & training
Hosting summary
Hosting depends entirely on which product is used. Catalophore Platform Pro runs on the customer's own site, with the web interface, scheduler, gateway, file server, licence authority, workers and database all installed locally and reference data pre-loaded, so no internet access is needed at runtime and fully air-gapped installations are supported. Catalophore Studio can likewise be deployed locally, on a shared server or on on-premise GPU hardware, and Innophore presents this as a guarantee that structures never leave the customer's environment. The CavitOmiX Copilot, by contrast, is operated by Innophore in a web environment whose jurisdiction is never stated: no hosting country and no region are named anywhere on the site. Internal compute is described as clusters of around 6,500 CPUs, with overflow onto Amazon Web Services and Google Cloud at peak times, again without a stated region. Encryption in transit and at rest is claimed. No sub-processor is named and no data processing agreement is offered.
Things to keep in mind
Risks and trade-offs to weigh before adopting Catalophore.
- Cavity matching produces hypotheses, not conclusions: the vendor states that AI is guidance and that the human remains the decision maker, and a ranked shortlist can easily be read as a verified result
- Reports look authoritative, with logos, UUID watermarks and statistics, which makes it tempting to circulate a run without re-examining the assumptions behind it
- Using the free PyMOL plugin sends the input structure or sequence to NVIDIA, Meta, DeepMind and the RCSB PDB, so confidential targets should not be pasted into it
- For the cloud copilot the vendor names no hosting country, publishes no sub-processor list and offers no data processing agreement, which is a real obstacle for a regulated programme
- Nothing documents whether uploaded data may be used to train models, and the phrase about improving the service is broad enough to worry a legal team
- Tokens are prepaid and non-refundable, so a misconfigured search that returns a large number of hits translates directly into spent money
- The partners and institutions listed on the site are collaborators and customers: their certifications, resources and results belong to them and should not be read as properties of the vendor
Setup & Integrations
Technical difficulty
Effort varies sharply by product. The CavitOmiX Copilot needs nothing beyond an online account. The PyMOL plugin is a manual archive install following the PyMOL Wiki procedure. Catalophore Studio ships as a container and can be deployed locally, on a shared server or on GPU hardware. Catalophore Platform Pro is the heaviest: one tarball and one Docker compose, then HPC clusters, Docker hosts and GPU nodes connected as Facilities, with worker health, mounts and licence quotas to monitor. Innophore says a single-team pilot can run within one sprint. Reading the output, however, requires structural biology expertise.
Deployment
Integrations
Supported languages
Behind Catalophore
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
What exactly does this entry cover?
How is a Catalophore search different from a sequence search?
What does it cost?
Is there a permanently free plan?
Is there an API?
Can the software run on our own infrastructure?
Who owns the data and the results?
Is customer data used to train models?
Does the vendor comply with the GDPR?
Who is behind Catalophore?
Should you pick Catalophore?
Catalophore is a specialist tool, and it makes no attempt to be anything else. Its value rests on one idea carried consistently across four products: comparing binding sites as physico-chemical point clouds instead of comparing sequences or folds. That idea has been peer-reviewed since 2014 and has since been developed alongside NVIDIA, Amazon Web Services and Google Cloud, whose scientists co-signed several of the papers.
What makes the offer credible for regulated research is the emphasis on explainability and control. Similarity formulas are published, matches carry a Gumbel e-value, property weights are adjustable, every run has a UUID recorded in an audit log, and exports use open formats. The on-premise platform can run entirely behind a firewall, air gap included, and Innophore states plainly that the intellectual property of uploaded data and of the results belongs to the customer.
The weaknesses are commercial and documentary rather than scientific. Only the token grid of the cloud copilot carries a public price, and even that is visible only after opening an account; the flagship products are quoted case by case. No data processing agreement, no sub-processor list and no hosting country are published, and nothing describes whether customer data may feed model training. The general terms date from 2018, are governed by Austrian law with the German version prevailing, and read like a software supply contract rather than a description of an online service. Several published links, including the terms PDF and the plugin archive, no longer resolve.
For a team working on enzymes, repurposing or off-target safety, the realistic way in is the free PyMOL plugin or the free Basic copilot, followed by a demonstration if the method proves useful.
- Choosing a selection results in a full page refresh.
- Opens in a new window.