Frequently asked questions

A scoping call, for which no data is needed. For a proof of concept, process data you already have is enough, as CSV or XLSX files or exports from your historian or LIMS, and we check together whether it is suitable before you commit. Most teams start with a retrospective study of one unit operation or one fermentation process.

In a proof of concept, the first results on your own data, structured and comparable datasets with first visualizations, typically arrive within one to three weeks. A predictive model you can plan experiments with follows within one to three months, depending on how much data you have and on the scope. A process development project with new experiments takes one to nine months, depending on the unit operation and the experimental effort.

There is no fixed number: it depends on how many factors you study, on your process and on how precise your analytics are. Studio trains a model on the data you already have and suggests the experiments that add the most information next, instead of a full predefined design. Across Novasign engagements this has cut experiments by 35–70% against each customer’s prior practice, and Bespark*bio reached the same process decisions from 8 runs instead of 22.

Yes. You need no central data platform, standardized data infrastructure or automation system, and most teams first use models to decide which experiment to run next and automate later, once they trust the model.

Yes, the proof of concept is optional. With a license you get Novasign Studio in the deployment you choose, an optional check of whether your data suits predictive modelling, an onboarding workshop on Novasign reference datasets and support from the Novasign team. After that, your team builds its own models.

In every phase, from early development, when only a few experiments exist, through late-stage process characterization to scale-up into pilot and manufacturing scale. Before a run you simulate to choose the conditions worth testing, and afterwards you bring the completed runs back to compare them and retrain the model. Where new runs are no longer possible, the model helps you understand the process from the data you have, for example to explain a deviation or to support the scientific rationale in your regulatory documentation.

Studio has dedicated process units for the bioreactor, for tangential flow filtration (TFF) and for chromatography. Other unit operations are modelled today on the modular engine or in a code node. Further units, such as refolding, precipitation, dead-end filtration, centrifugation and formulation, are added to the interface once they are tested.

Novasign Studio has been applied to microbial fermentation with E. coli and yeasts such as Komagataella phaffii, to mammalian cell culture with CHO and HEK cells, to insect cells, to stem cells and to Vero cells in a viral vector vaccine process. The products range from monoclonal and bispecific antibodies, antibody-drug conjugates, fusion proteins, enzymes and biosimilars to viral vectors and AAV, cell therapy, mRNA and plasmid DNA, precision fermentation, cultured food and microalgae. Studio is not limited to these: the model adapts to your organism, product and process, so other hosts and products need no pre-built template.

No. Process scientists and engineers are the primary users, and the core features work through a graphical interface without programming. Data scientists can add their own Python models and scripts in a code node, which colleagues then use without writing code.

Yes. Every data processing step, training decision, simulation and recommendation is visible in the workflow, so your team can inspect and document the path from raw data to result. That lets you justify a process decision to a client or within your own organization.

A trained model of your process with all its stages and its complete training history. You can use it to simulate new conditions, compare scenarios and plan the next experiments, and it gives the same predictions every time it is restored.

Yes. Novasign Studio runs on your own servers, in your own cloud account or on an appliance that Novasign builds and ships, so you decide where your process data is stored. A managed Novasign Cloud, run by Novasign, is rolling out.

You do. All models, workflows and results you build in Novasign Studio belong to your organization, and Novasign owns only the platform’s own components. Your data is not used by Novasign for any other purpose.

No. Novasign runs a formal compliance program covering ISO 27001, SOC 2, GDPR and the EU AI Act and is working toward ISO 27001 certification. The Trust Center at trust.novasign.at shows the program as evidence is collected, and security questionnaires are answered on request.

Yes, with the on-premise and appliance options. Studio then has no public internet exposure, keeps metrics and logs inside your environment and installs from an offline bundle with its own private registry. The fully disconnected setup and its update path are planned with your team for each installation. Sign-in then connects to an identity provider inside your network, such as Active Directory, LDAP or AD FS.

No. Each sponsor sees only the data, models and workflows of their own program, and access is set at company, site, department and user level.

Inside Novasign Studio. You share selected experimental and modelling results with a client, who explores them there instead of receiving slides or spreadsheet exports.

For any bio-based production process, from mammalian cell culture such as CHO, cell and gene therapy and mRNA to microbial and precision fermentation and microalgae. You describe the vessel, its components, the streams in and out and the parameters that drive the kinetics, and that description stays the same whatever the host and product.

The course of every component you measure, such as biomass, substrates, product titer and quality attributes like a glycoform or a charge variant. From these it derives performance metrics such as volumetric productivity, space-time yield, integral viable cell count (IVCC), doubling time and purity.

Batch, fed-batch with any feed profile including feedback-controlled feeding, semi-continuous operation such as fill-and-draw or repeated fed-batch, and continuous operation with or without cell retention, from perfusion to a chemostat or turbidostat.

The one that answers your question with the runs you have. A fractional factorial screens many factors in few runs, and a central composite design studies the curvature of fewer factors. If you are unsure, describe your process and goal to the AI Assistant, which proposes a design and explains why it fits, or work it out with a Novasign modelling specialist.

Yes. Enter them as your own run table with the levels you actually chose and combine them with the design you generate next. The model trains on both.

Studio suggests it, and you decide. The model predicts the response at conditions you have not run, ranks the predictions inside your limits by your objective and suggests the best as next experiments, which you can inspect and change before anything is booked. If you prefer, Novasign runs this loop with you as a process development project.

In Novasign projects, two to three iterations are typical, depending on how well the first model covers the region you want to operate in. If the confirming run agrees with the prediction, one can be enough. If it does not, the model is retrained on the new run and becomes more accurate exactly there.

Yes, if your process returns to a stable state after a setpoint change within a run. Studio then turns the existing design into multi-stage runs within your constraints: how many stages each run may have, how many runs you can book and which factors are safe to change during a run.

On the publications page, which lists Novasign’s peer-reviewed papers and preprints in process modeling, each with its DOI. All but one of the peer-reviewed papers are open access, and nothing on the page sits behind a form.