Page 85 - IJAMD-1-1

International Journal of AI
for Material and Design MMDB: A comprehensive biofabrication database

On completion of data extraction, we employed a and insights is paramount. This demand has led to the
gradient boosting decision tree-based multi-output development of tools that distill complex information
regressor from the sci-kit-learn library. to fit the into actionable knowledge. One such essential tool is
29
relationship between organ type, cells, biomaterials, and our knowledge service component, a critical part of our
the four printing parameters. Consequently, we can specify database, which comprises two principal parts: knowledge
a particular organ, cell, biomaterial, and printing method analysis and knowledge mapping of organ functions. The
to derive the most suitable printing parameters under knowledge analysis section delves into the examination
those conditions. The optimization of printing parameters of research trends, research hotspots, and international
holds the potential to deliver significant time and cost collaboration. The exploration of research trends
efficiencies for researchers engaged in the construction of illuminates the developmental trajectory of specific sub-
tissues or organs through 3D bioprinting. topics in biofabrication, such as publication and citation
For the formulation of the second model, it was not patterns. The identification of research hotspots focuses
necessary to limit our focus to articles utilizing 3D bioprinting. on prevalent topics in biofabrication, such as clustered
Instead, we considered all types of manufacturing strategies research topics and the top ten most prolifically published
encompassing 2D culture, spheroid-based, scaffold- journals. The examination of international collaboration
based, and organ chips. Furthermore, the extraction of quantifies the frequency of cooperation between different
finer-grained fabrication parameters was required. We nations and research topics.
individually extracted parameters from the three core In the domain of biofabrication research trend analysis,
components involved in the in vitro construction of human our methodology commenced with the extraction of
organs, namely, cell parameters, biomaterial parameters, research paper topics, discerned either from the abstracts
and culture platform parameters. or author-provided keywords. In instances where keywords
Specifically, cell parameters included aspects were explicitly furnished by the authors, we posited these
such as cell type and seeding density. Biomaterial as representative of the paper’s thematic focus. Conversely,
parameters incorporated details such as biomaterial for documents devoid of author-specified keywords, we
type, concentration, and modifications. Culture platform utilized the pre-trained language model KeyBERT. to distill
31
parameters varied according to the manufacturing strategy keywords and key phrases from the abstracts, subsequently
employed. For instance, in the scaffold-based culture of considering these as the paper’s topics. Subsequent to
tissues or organs, factors such as scaffold porosity, diameter, topic extraction, we applied the K-means algorithm. to
32
and thickness were considered. In the case of organ chips, categorize the topics into distinct clusters. We determined
manufacturing parameters such as chip channel width, type the preeminent topic within each cluster by consulting
of serum used, and whether the chip was self-circulating expert assessments and appraising the frequency of each
were considered. Following this, it was necessary to topic’s occurrence. Quantifying the volume of publications
extract functional indicators of the organs. For instance, per topic enabled an analysis of publication trends over
for the liver, we utilized image recognition techniques to time to elucidate the prevailing research trajectory within
extract quantities of albumin and urea secretion from the the field of biofabrication. Furthermore, we scrutinized the
figures in the included articles over given days. Once all citation trends of these publications to gauge the topics’
parameters were extracted, we employed the XGBoost academic impact. In addition, we assessed the periodical
regressor. to establish the nonlinear relationship between and institutional dissemination patterns of the publications
30
these fabrication parameters and functional indicators. In pertinent to each topic.
simple terms, given a set of fabrication parameters, the With respect to the analysis of research hotspots, we
model can predict functional indicator values for specific clustered the topics of the selected research papers and
days. This model has the capability to simulate potential tabulated the number of publications within each cluster.
functional secretion values without the need for wet-lab
experiments, thereby providing important references for The size of a cluster was indicative of the topic’s research
the experimental design of biofabrication researchers. intensity. Special attention was accorded to topics that
This invaluable tool could significantly streamline research demonstrated a rapid escalation in publication volume in
processes, reduce costs, and promote the rapid progression recent years, marking them as nascent frontier research
of advancements in the biofabrication field. areas. In addition, we identified the top 10 international
journals and institutions that made substantial
2.4. Knowledge service contributions to the field.
In the constantly evolving landscape of biofabrication Regarding the analysis of international collaboration,
research, the need for comprehensive understanding we interpreted the keywords and key phrases of the papers

Volume 1 Issue 1 (2024) 79 https://doi.org/10.36922/ijamd.2420