Skip to main content

Bisy increased enzymatic activity 2.4-fold with model-based DoE

In two rounds of designed experiments, Bisy increased the enzymatic activity of a Komagataella phaffii (formerly Pichia pastoris) process for an industrially relevant enzyme 2.4-fold over its starting process.

About Bisy

Bisy is a biotech company working in the field of enzymes and recombinant proteins. Its upstream work is microbial expression in yeast. Bisy optimized one of those upstream processes with Novasign.

The project partner was the Universitat Autònoma de Barcelona (UAB), through its Department of Chemical, Biological and Environmental Engineering.

The objective

The goal was to raise the enzymatic activity of the upstream process for an industrially relevant enzyme made in Komagataella phaffii beyond Bisy’s starting process. Each experiment is a fermenter run, so every run has to tell Bisy something new.

Model-based experiment design

A set of experiments was set up and then refined with model-based design of experiments in Novasign Studio. The model trained on the runs so far predicts how conditions nobody has run would behave, and the most promising ones are performed next.

The campaign ran in two rounds of 16 parallel cultivations over three process parameters, which stay confidential, and two substrates. The first round combined low, medium and high settings of each parameter on both substrates. The second round, a model-based DoE, used one of the two substrates and moved into a new region of the design space. The runs of both rounds then trained a hybrid model that predicts the process under new conditions and supports scale-up.

Design space of the two rounds
  • First DoE
  • Second DoE, model-based
  • Best run of each round, as a fold of the starting process
  • From the first round’s best run to the second’s
The 32 runs of two rounds of designed experiments as points on three coded parameter axes. The 16 runs of the first round spread over the low, middle and high setting of each parameter, including runs with a high third parameter. Its best run, with 1.5-fold the enzymatic activity of the starting process, sits on the edge of that design, at the low first, high second and lowest third parameter. The 16 runs of the second round all sit at the lowest third parameter and move past that edge to lower values of the first parameter and higher values of the second. Their best run, with 2.4-fold that activity, lies furthest out in that direction, and an arrow joins the two best runs.
RunParameter 1Parameter 2Parameter 3
1.1-1-2-1
1.2-1-21
1.3-1-1-2
1.4 (1.5×)-11-2
1.50-2-1
1.60-20
1.70-20
1.80-21
1.90-1-2
1.1000-2
1.1100-2
1.1201-2
1.131-2-1
1.141-21
1.151-1-2
1.1611-2
2.1-2.52-2
2.2-21-2
2.3-22-2
2.4 (2.4×)-23-2
2.5-10.5-2
2.6-11-2
2.7-11-2
2.8-12-2
2.9-12-2
2.10-13-2
2.11-13.5-2
2.1201-2
2.1301-2
2.1402-2
2.1503-2
2.160.52-2
The best run of the first round sat on the edge of that round’s design: low Parameter 1, high Parameter 2, lowest Parameter 3. The model-based second round extended the design past that edge, and its best run reached 2.4-fold the enzymatic activity of the starting process.

Benefit for Bisy

The first DoE round increased enzymatic activity 1.5-fold over Bisy’s starting process. The model-based second round took it to 2.4-fold, and every one of its runs beat the starting process. Beyond these figures and the design itself, the run data stays with Bisy.

Stage · Enzymatic activity
Three bars of enzymatic activity relative to the starting process: the starting process at 1.0, the first DoE round at 1.5-fold and the model-based second round at 2.4-fold.
StageEnzymatic activity (×)
Starting process1
First DoE round1.5
mDoE roundmodel-based DoE2.4

We reduced development timelines by more than 50% in an initial proof-of-concept project, using Novasign Studio to accelerate process development for new products.

Anton Glieder, CEO, Bisy GmbHBisy GmbH

Bring a process you are optimizing now

Bring it to a demo, and we will build a model against its own run table.

Book a demo
Related