FAQ

The typical timeline is about 8–10 months for antibodies and 10–12 months for enzymes.
This may vary depending on the target protein and assay methods, but we offer flexible support based on project needs.

aiProtein typically provides around 10 candidate sequences.
These candidates are designed to ensure sequence diversity, allowing you to select the final candidates based on your own evaluation criteria. Having multiple sequences can also support intellectual property (IP) strategies.

RevoAb, on the other hand, provides up to three candidate sequences.
For RevoAb only, there is no charge if no candidate sequences can be delivered.

No.
Providing experimental data or proteins is not required.
However, sharing available data or proteins may help improve the efficiency of sequence space design and evaluation. The service can still be carried out even if they are not available.

No.
There is no significant difference.
Regardless of whether experimental data is provided, the project follows RevolKa’s standard experimental and analysis workflow, so the overall timeline remains largely the same.

Yes.
In general, RevolKa manages the entire process from design to protein expression and evaluation.
However, roles can be shared with you depending on the project timeline and evaluation requirements.

RevoAb is a fully online antibody sequence optimization service.
It aims to improve developability properties such as yield and stability. By submitting your antibody sequence through our online form, our proprietary algorithm generates up to three promising variant sequences.

Results may be delivered as quickly as the same day, depending on the request.

RevoAb represents the first step of antibody optimization using aiProtein and provides a simplified version of sequence space design.
For antibodies that show promising results with RevoAb, you can move to the full aiProtein service, which enables further functional improvements through advanced machine learning.

In principle, all intellectual property generated through this service belongs to the customer.

Due to confidentiality agreements with our clients, we cannot share detailed information about domestic projects.
A summary of our experience will be published soon.

Yes.
We currently have international projects that are ongoing or under discussion, and we plan to further expand our global activities in the future.

Yes.
We have experience in modifying enzyme substrate specificity.
Enzymes typically exhibit high diversity in amino acid sequences and functional properties, making model development challenging. However, because RevolKa’s technology learns from experimental data, we can address property modifications that are difficult to achieve with conventional approaches.

Yes.
RevolKa leverages machine learning to identify solutions that are difficult to achieve using conventional rational design or simple optimization approaches. In some cases, variants with improved functionality identified by machine learning can later be rationally explained.

Please contact us to discuss project feasibility.

Using RevolKa’s proprietary sequence space design technology, we first narrow down candidate residue positions and amino acid substitutions. Mutations are then introduced randomly within this defined sequence space.

In many cases, the desired results can be achieved in a single design and evaluation cycle.
If the results do not meet your expectations, multiple cycles can be carried out for further optimization.

Yes.
RevolKa designs based on the full-length sequence of the target protein, allowing interactions between distant residues to be considered. This enables the identification of optimal mutation combinations that are difficult to find using local design approaches.

In principle, machine learning is challenging if no improved variants are present in the training data.
However, even a small proportion of improved variants (a few percent) is sufficient, and negative data also provide valuable information for optimization.

Thanks to RevolKa’s sequence space design technology, we have not encountered cases where improved variants could not be obtained. We also establish a milestone after training data generation to allow you to decide whether to proceed.

No.
Antigen sequence information alone is not sufficient.
RevolKa uses experimentally measured cross-reactivity data for the starting antibody and its variants, and applies machine learning to optimize performance. This enables balanced binding and function across both mouse and human.

Around 100 variant data points are typically required.
The actual number may vary depending on the target protein and evaluation criteria. If needed, machine learning can begin with a smaller dataset to meet project timelines.