Page 67 - IJAMD-1-2

International Journal of AI for
Materials and Design

ORIGINAL RESEARCH ARTICLE
Machine learning-driven prediction of gel

fraction in conductive gelatin methacryloyl
hydrogels

Xi Huang* , Ye Xuan Wong, Guo Liang Goh , Xinchao Gao , Jia Min Lee ,
and Wai Yee Yeong*
School of Mechanical and Aerospace Engineering, Nanyang Technological University, Singapore

Abstract
Gelatin methacryloyl (GelMA) hydrogels, combined with conductive fillers like Poly(3,4-
ethylenedioxythiophene) polystyrene sulfonate (PEDOT:SPSS), present significant
promise for tissue regeneration due to their biocompatibility, biodegradability, and
electrical conductivity. However, optimizing the curing process of the hydrogel
is challenging due to a lack of an existing model for gel fraction prediction. This
complexity is further heightened when additional variables such as bioink formulation
and crosslinking parameters are considered. This study leverages machine learning
(ML) to predict the gel fraction of GelMA-PEDOT:SPSS hydrogel based on the
combination of three types of features: Bioink formulation, crosslinking parameters,

*Corresponding authors: and absorption coefficient. The two key objectives of this study are to develop an ML
Xi Huang model to predict gel fraction from bioink formulation and crosslinking parameters
(huang.xi@ntu.edu.sg) such as ultraviolet (UV) power intensity and UV irradiation duration, and to create
Wai Yee Yeong
(wyyeong@ntu.edu.sg) an ML model to predict gel fraction through the absorption coefficient instead of
crosslinking parameter. In the first ML model, support vector regression achieved
Citation: Huang X, Wong YX, the highest accuracy with a mean absolute percentage error (MAPE) of 3.13% and
Goh GL, Gao X, Lee JM, Yeong
WY. Machine learning-driven an R² of 0.79. This model allows the user to select optimum bioink formulation
prediction of gel fraction in and crosslinking parameters to achieve the required gel fraction with minimal
conductive gelatin methacryloyl experiment. For the second ML model that utilizes a combination of absorption
hydrogels. Int J AI Mater Design.
2024;1(2):3807. coefficient and bioink formulation, deep neural network models achieved a MAPE
doi: 10.36922/ijamd.3807 of 6.31% and an R² of 0.54. The absorption coefficient model shows promise for a
Received: May 31, 2024 non-destructive, real-time assessment of gel fraction, enabling more precise control
over the hydrogel properties during the curing process. These results demonstrate
Accepted: July 12, 2024
ML’s capability to efficiently optimize hydrogel formulations, significantly cut down
Published Online: August 8, 2024 experimental efforts, and improve precision in 3D bioprinting and other hydrogel
Copyright: © 2024 Author(s). applications, thereby advancing the field of tissue regeneration.
This is an Open-Access article
distributed under the terms of the
Creative Commons Attribution Keywords: 3D bioprinting; 3D printing; Biofabrication; Machine learning; Hydrogel;
License, permitting distribution, Composite
and reproduction in any medium,
provided the original work is
properly cited.
Publisher’s Note: AccScience
Publishing remains neutral with 1. Introduction
regard to jurisdictional claims in 1-3
published maps and institutional Hydrogels have wide usage in 3D bioprinting application. Among these, gelatin
affiliations. methacryloyl (GelMA) was highly valued due to its biocompatibility, biodegradability,

Volume 1 Issue 2 (2024) 61 doi: 10.36922/ijamd.3807