Page 100 - AJWEP-v22i3

Sonsare, et al.

rather than relying on a centralized cloud infrastructure. environmental modeling and simulations. It discusses
The research investigates the integration of artificial the systematic process required for successful
intelligence with edge computing, emphasizing its application management, involving expertise in various
prospects for immediate data analysis and decentralized scientific domains. The study systematically reviews
54
decision-making in health care, smart cities, industrial tools for maximizing biological information of genes,
automation, and autonomous systems. It emphasizes the summarizing over 300 tools, databases, and algorithms
importance of compact artificial intelligence models and for analyzing differentially expressed genes. It provides
strong security mechanisms. ORCAN is an internet- guidelines to assist researchers in effectively mining
48
based meta-server for single-click protein sequence gene functions and interactions. The review highlights
labeling that increases sensitivity and accuracy by 1 – trajectory inference tools such as Monocle, Slingshot,
2% while correcting conflicting orthology predictions. and scVelo, focusing on their unique features and
49
The study explores the application of knowledge graph applications. It discusses gene-phenotype association
embedding (KGE) models in biological systems, analysis methodologies, including FUSION and
highlighting their ability to represent complex biological PrediXcan, to uncover genetic variations linked to
knowledge as graphs. KGE models demonstrate superior diseases. 55
predictive accuracy and scalability in tasks such as Another evaluation investigates sludge extract
predicting drug-target interactions and polypharmacy management, with an emphasis on resource recuperation
side effects. The study discusses the challenges of data and reduction of environmental impacts for zero-
quality and interpretability associated with KGE models waste discharge. It addresses the issues, constraints,
while emphasizing their potential in various biological and solutions for maximizing reuse for ecological and
applications, including genomics and proteomics. regulatory adherence. Badawi et al. investigate the
56
Overall, KGE models are positioned as a promising tool
for advancing biological data analysis and predictive application of orange peel-derived activated carbon
modeling. 50 supported by cobalt ferrite as a ferromagnetic scrubber
An additional study criticizes the present default for the treatment of effluent from pulp and paper mills.
assumptions regarding product consumption in Its promise as an environmentally friendly approach is
Registration, Evaluation, Authorisation, and Restriction indicated by the findings, which demonstrate excellent
of Chemicals standards for chemical emissions, rates of pollution removal. 57
claiming that these values are excessively cautious and
geographically constrained. The project investigates 3. Methodology
51
the potential of metabolomics in drug development,
focusing on its applications in clinical pathology, 3.1. Dataset preprocessing
biomarker identification, and metabolic subtyping. There are 188 molecular graphs in the MUTAG
It focuses on the application of machine learning dataset. Atoms are shown as nodes, and chemical
methods to analyze complicated metabolic data. The linkages as edges in a graph that represents each
52
study proposes a machine learning-powered Perturb molecule. Carcinogenic (1) and non-carcinogenic
and Observe (P&O) algorithm, named artificial neural (0) labels are applied to each graph. Normalizing the
network+P&O, enhancing conventional methods for node characteristics (atom types) to a range of 0 – 1
maximum power point tracking. The artificial neural is the first step in preprocessing the dataset. After
network model predicts the duty ratio, improving that, the dataset is divided into testing (20%) and
convergence speed and energy yield compared to training (80%) sets. G = (V, E) is the representation
traditional P&O algorithms. The proposed algorithm of each molecular network in the dataset, where V
effectively reduces the number of iterations needed stands for the set of nodes (atoms) and E for the set
to reach the maximum power point, ensuring optimal of edges (bonds). Each node represents atom types
performance. 53 as one-hot vectors, and characteristics such as bond
An additional study focuses on managing compute- types (single, double, etc.) are linked to edges. To
intensive applications for high-performance systems guarantee uniform feature ranges, node features were
to enhance operational effectiveness. It highlights normalized using min-max scaling. An adjacency
the challenges in designing and deploying high- matrix A was constructed for each network to encode
fidelity applications for domain experts. The study connections. If nodes I and J are linked, then Aij = 1,
emphasizes the importance of computational power in and if not, Aij = 0.

Volume 22 Issue 3 (2025) 94 doi: 10.36922/AJWEP025070041