Page 68 - IJAMD-1-2

International Journal of AI for
Materials and Design
Machine learning for gel fraction prediction

and tunable mechanical properties. These characteristics in determining the DoC of the sample, but may disrupt
25
4,5
make GelMA an excellent candidate for creating extracellular the cells in bioprinting. Most of the in situ measurement
matrix in the field of tissue regeneration. Electrical signals setup is too bulky to be integrated in a 3D printer or
are crucial to the electroactive tissue in the human body, could be destructive to the bio-sample. Besides, although
such as heart, muscles and nerves. The cells are highly these methods can help with the in situ measurement of
9
6
7,8
regulated by affecting the intracellular signaling pathways as the crosslinking process, no prior work has been done on
well as intracellular microenvironment. Therefore, the use gel fraction prediction from the bioink formulation and
10
of conductive hydrogel for tissue regeneration will facilitate crosslinking parameters to estimate the gel fraction before
the transmission of electrical signals, thereby promoting cell the experiment. There is a need for a model capable of
communication, proliferation, and differentiation within predicting gel fraction based on the parameters before the
the engineered tissue. 11,12 Poly(3,4-ethylenedioxythiophene) experiment, and a model for an easier in situ measurement
polystyrene sulfonate (PEDOT:SPSS) is known for its of the gel fraction.
biocompatibility, solubility, excellent electrical conductivity, Prediction and control of the gel fraction are of
and mechanical flexibility. 13-15 The integration of GelMA critical importance in tissue regeneration, particularly
with PEDOT:SPSS opens up new avenues for developing following the fabrication of engineered tissues and their
multifunctional biomaterials that leverage the strengths of subsequent transplantation into in vivo environments. The
both components. 16,17 crosslinking parameters, including ultraviolet (UV) light

Adding conductive fillers such as PEDOT:SPSS into power and exposure duration, significantly influence the
GelMA hydrogels significantly affects the curing process. 18,19 gel fraction of the crosslinked hydrogel. On the one hand,
The incorporation of these fillers can enhance the electrical if the engineered tissue is not fully cured, it will degrade
conductivity of the hydrogel, which is beneficial for more rapidly than anticipated. On the other hand, if the
applications involving electroactive tissues. 20,21 However, engineered tissues are overcured, there will be a dimensional
this enhancement comes with a tradeoff. The presence of mismatch between the designed structure and the actual
conductive fillers can impede light transmission through tissue, which is especially problematic for certain structural
the hydrogel, affecting how far the light can penetrate. 16,22 features, such as narrowed channels within the engineered
Since the curing process of GelMA often relies on scaffold. This dimensional mismatch could impede the
photoinitiated crosslinking, reduced light penetration can migration of cells into these channels, thereby hindering
lead to incomplete or inefficient curing, compromising the the tissue regeneration process. While increasing the UV
26
mechanical properties and functionality of the hydrogel. energy absorbed by GelMA may enhance the mechanical
The balance between conductive filler concentration properties by achieving higher crosslinking, this is not
and curing effectiveness is complex and requires careful always ideal. A high mechanical strength often compromises
optimization. Increasing the amount of conductive cell viability and proliferation, 27-29 and the excessive UV
filler not only improves electrical conductivity but also exposure can be detrimental to the encapsulated cells
decreases light transmission, necessitating a precise in GelMA. 28,30-32 Furthermore, each component of the
balance to achieve optimal curing efficiency. This delicate bioink, such as the concentrations of GelMA, lithium
interplay between filler concentration and curing efficacy phenyl(2,4,6-trimethylbenzoyl) phosphinate (LAP), and
underscores the need for a comprehensive understanding PEDOT:SPSS, affects the crosslinking rate, adding an
of the curing dynamics and the material properties additional layer of complexity. Generalizing a model for
influenced by different filler concentrations. the gel fraction thus requires analysis of an enormous
Conventionally, the optimization for the crosslinking dataset, highlighting the difficulty of balancing multiple
of hydrogel is done by performing experiments on a parameters to achieve the desired hydrogel properties.
wide range of parameters, which is costly in terms of the Machine learning (ML) techniques offer significant
materials used and time-consuming. The experimentation potential in optimizing both the formulation of GelMA-
cost can be reduced by having a model to predict the gel PEDOT:SPSS hydrogels and the bioprinting process
fraction beforehand, while the precision of the crosslinking itself. ML algorithms can analyze vast datasets to identify
process can be improved with in situ characterization patterns, predict outcomes, and optimize parameters more
technique by fine-tuning the curing process in real efficiently than traditional methods. 33-39 ML is widely used
time. 23-25 There is an attempt on quantifying the degree of in material science for characterization and optimization
conversion (DoC) of photopolymerizing resin from the of synthesis processes. 40-42 By utilizing ML, researchers can
in situ measurement of the sample’s refractive index, but explore the parameter space, optimize the formulation, and
24
a bulky setup for the microscope system is required. For predict the performance of the hydrogels under various
vat polymerization, in situ ultrasonic monitoring can help conditions. 43,44 In addition, ML can be applied to optimize

Volume 1 Issue 2 (2024) 62 doi: 10.36922/ijamd.3807