Skip to content
Matmerize

Design of PHA-based Polymers with
Attractive Mechanical Properties

Given the urgent need for biodegradable polymers to
combat plastic pollution, it is crucial to develop
sustainable materials that not only possess suitable
functionality, such as appropriate mechanical
properties, but also match or exceed the performance
of their non-sustainable counterparts. If these
materials are further limited by specific chemistries,
such as biodegradable polyhydroxyalkanoates (PHAs),
Matmerize’s PolymRize™ platform can help.
Leveraging artificial intelligence-based
methodologies, PolymRize™ is designed to create
such essential polymers. In fact, we have helped
companies like Kimberly-Clark and CJ
Biomaterials to do just that.

Kimberly Clark Collaboration

Kimberly-Clark has an ambitious strategy for 2030 and new sustainability goals for the company that include a
global ambition to improve the lives and well-being of 1 billion people in underserved communities around the
world by 2030, with the smallest environmental footprint. To achieve this, Kimberly-Clark and RWDC Industries
collaborated with Matmerize and used the PolymRize™ platform.
In this project, an initial dataset of 150 data points was used to build
predictive models to screen over 20,000 polymer formulations by
turning various knobs in the formulation. A dozen new formulations
were identified that were predicted to exceed the target performance
variable by 30%, as shown in Figure 1. The model was able to accurately
mirror known physics; it correctly learned which knobs will affect the
target variable. PolymRize™ was thus able to sweep massive swathes of
the search space in seconds. The ability to virtually screen tens of
thousands of polymer formulations at minimal cost, and at scale, is
necessary to access and assess the breadth of the chemical-process space.
Multiple properties, multi-property learning, automatic generation of
candidate materials and multi-property optimization can be handled in
PolymRize™ using proprietary algorithms [1]. Such a capability can not
only be used to streamline R&D efforts by intelligently selecting the “next
best experiment” but can also be used for “active learning” (Figure 2) to
efficiently converge on to the target materials in an accelerated manner.
“ The Matmerize Team has done a phenomenal job not only in assisting Kimberly Clark to identify novel polymer solutions
but also in improving our internal AI capabilities through workshops and readily available technical support. ”
– Bryan Haynes, Technical Director, Global Nonwovens, Kimberly Clark –
Biodegradable Polymer Design
Enabled by PolymRizeTM
Renewable &
sustainable carbon
feedstocks
Fermentation
Plants convert CO2 via
photosynthesis
Downstream
processing
Plastic wastes undergo
biodegradation
Product use and
disposal
AI-assisted design
& selection of PHA
polymers
PHA polymers with
desired target
properties
Figure 1. Histogram of target performance
variable for several formulations.

Active Learning with Small Datasets:

Collecting accurate experimental data is
expensive and time-consuming for
organizations. PolymRize™ has the capability
to build custom models using active learning
by extracting valuable information from small
datasets. These models continuously improve
in performance as they iteratively incorporate
new predicted values, thus reducing the time
and resources needed to achieve
biodegradable design goals.

CJ Biomaterials Collaboration

The collaboration between CJ Biomaterials, a division of South Korea-based CJ CheilJedang, and Matmerize was
aimed at advancing sustainable polymer solutions to combat plastic pollution. CJ Biomaterials, a leading producer
PHA biopolymers, utilized Matmerize's PolymRize™ platform to optimize bio-based polymers. The platform
helped CJ Bio ensure that these bio-based polymers met specific performance criteria across various applications.
CJ Biomaterials developed PHACT, a 100% bio-based PHA created through an innovative fermentation process.
This biopolymer degrades naturally in soil and ocean environments, significantly reducing the harmful impact of
plastic pollution on marine ecosystems. One of the significant challenges CJ Bio faced during development was the
time-consuming and resource-intensive process of characterizing core properties. By leveraging Matmerize's
custom model training and predictive analytics platform, PolymRize™, CJ Biomaterials could quickly estimate the
performance of PHACTs. This capability enabled faster decision-making and reduced time and costs compared to
traditional methods.
[1] Christopher Kuenneth, Arunkumar Chitteth Rajan, Huan Tran, Lihua Chen, Chiho Kim, and Rampi Ramprasad, Polymer informatics with
multi-task learning, Patterns, 2, 4, 100238 (2021).
Progressive improvement of
model performance
Strategized design for the
next experiments
Reduced time and cycle
to achieve the design goal
ACTIVE
LEARNING
CYCLE
SMALL
DATASET
TRAIN
CUSTOM
MODEL
AUGMENT
NEW DATA
DESIGN
CANDIDATES
VALIDATE
Figure 2. Active learning to accelerate formulation/process design, start from
small data.