Traditional protein engineering is slow and costly
Anyone who optimizes proteins using standard methods such as directed evolution or deep mutational scanning knows how time-consuming the process is. Development takes a long time, costs a lot of money and in the end you can’t be sure if you have actually achieved the best result. This is where Exazyme comes in.
With Exazyme you will reach your goal faster and cheaper
The Exazyme algorithm will get you to your goal faster, with significantly fewer experiments, and thus more cost-effectively. Instead of having to do a lot of experiments yourself, Artificial Intelligence predicts exactly the attributes you need based on amino acid sequences. All you have to do is apply the predicted sequences in the lab and evolve the protein using the algorithm. Through our algorithm you will find the highest absolute values available, because it is designed to search the entire search space – this is hardly possible with directed evolution and other standard methods. In addition, our algorithm also includes bad values in order to make better predictions. So Exazyme is your path to the best protein development.Book conversation
With our algorithm we support many companies and scientists from different industries.
16xfewer experiments than with standard procedures.
FastResults through automation.
20Data points are sufficient for the algorithm.
10free suggestions for new customers.
With Exazyme you will benefit from many advantages that will advance your work.
Fast to the best result
With Exazyme, you need to perform 16 times fewer experiments than with standard methods. This saves an enormous amount of time. In addition, our algorithm allows you to find absolute values that are difficult or impossible to achieve with other methods.
Since you need to perform fewer experiments with Exazyme, you not only save valuable time, but also automatically save money. In addition, you need to use fewer working materials.
No large data sets
The algorithm already works from two data points. So there is no need for large data sets. Before the actual start, it tests whether the measured values are actually sufficient. This way you can be sure that the algorithm really works.
Case studies from our customers
ProblemExazyme's technology should be used to improve the activity of a carboxylase enzyme. Using conventional methods such as Directed Evolution, an improvement of 500x in activity was achieved after 6 years compared to wild type.
SolutionAfter only one iteration, our AI technology was able to produce a variant that was 3x better than the current best at 500x.
ProblemExazyme should be used for a biocatalyst for fuel cell reaction.
DestinationThe goal was to significantly speed up the experiments and reliably predict effects of mutations.
SolutionUsing just two data points, Exazyme was able to predict the effects of mutations. In addition, measurement data for one family of hydrogenases allowed experiments for another family to be accelerated 10-fold.
ProblemThe thermostability of a phytase enzyme should be further improved after long experiments with directed evolution.
SolutionAfter just one iteration, our AI technology was able to generate a variant that produced a 1 degree Celsius improvement in thermostability.
ProblemOur system should be used for a carboxylase. A biocatalyst for CO₂ reactions should be developed.
SolutionNine out of ten exazyme proposals in an experimental round showed significant activity. With error prone PCR, it ended up being two out of ten variants.
ProblemExazyme should be used for an imine reductase in pharmaceutical production.
DestinationOur system should ensure that significantly fewer variants are needed. The process should be faster and more effective.
SolutionThe result: With Exazyme, our customer needed 16x fewer variants compared to directed evolution.
Exazyme wants to move society forward. Behind the idea are biochemist Dr. Jelena Ivanovska, AI expert Dr. Ingmar Schuster and engineer Philipp Markert. Our common goal is to significantly accelerate protein discovery and design, and to help scientists deliver better biocatalysts. This is because while protein-based catalysts are playing an increasingly important role in the sustainable chemical industry and the reuse of CO₂, protein receptors and artificial antibodies are important in medicine.
Exazyme continues to develop from day to day. To make our system even better, we collaborate with recognized scientists and experts in the fields of biocatalysis and AI.
You are interested in our work, but cannot yet imagine how exactly our cooperation will work? We show you.
Answers to the most frequently asked questions
Many questions come up again and again in our work. We would like to answer some of them here in a very straightforward way.
Changing the amino acid sequence of a protein typically also changes its three-dimensional structure. However, all the information needed is already contained in the AA sequence, so the AA sequence alone can be used to predict the three-dimensional structure. If one enters another structure into the AI algorithm, then this neither brings new, statistical information nor a better prediction. Furthermore, not all proteins have a fixed structure. Thus, since our algorithm relies solely on the AA sequence, it works without secondary structures.
This depends on several points - but above all on the quality of the data. But: in many cases, only two data points were needed in the end to figure out which candidates to prioritize in experiments. To test whether the algorithm works, you basically need 20 data points.
When 20 data points are available, we test whether the algorithm works in your case. If our test is positive, we guarantee a strong acceleration compared to standard methods like directed evolution or deep mutational scanning.
While we can't give you an exact number, we can say that fewer synthesized sequences are needed compared to standard methods. For AI, more repetitions with smaller size are beneficial.
In the vast majority of cases, the algorithm works. Nine out of ten users for whom the test was negative found out in the end that there was a measurement problem. When they finally improved their measurements, it turned out that the system worked after all.
First, simply upload your measurement data to our web app. The app automatically checks if there is enough data to make a good prediction. If sufficient measurement data is available, you can make various configurations. You can choose between random mutations, digital deep mutation scans, a fixed candidate list and other options. The AI now predicts which sequences would improve protein properties or provides information about protein quality. You can easily download the corresponding sequences from our web app in the form of a document and use them immediately. You simply use the suggested sequences in your lab tests, improving the protein while producing more important data for the algorithm. Finally, if necessary, you can repeat the process as many times as you like to keep improving the protein.