Page 99 - AJWEP-v22i3

Advancing molecular property prediction using graph neural networks

that improves metabolite identification in unfocused table are two examples of how the update improves
metabolomics without the need for a full spectrum data display. It harbors a wider variety of organisms,
library. MetDNA utilizes initial seed metabolites and such as bacteria, fungi, and vertebrates. For local
their reaction-paired neighbors to expand annotations, analysis, users can utilize a standalone version or
achieving approximately 2,000 metabolite annotations upload datasets. Updated to version 5.0 with larger
43
from a single experiment. The methodology allows genome sets, eggNOG is a publicly accessible database
for quantitative assessment of metabolic pathways and for orthology connections and functional annotations.
supports integrative multi-omics analysis. The study A total of 4.4 million orthologous groups from 379
demonstrates the algorithm’s effectiveness across taxonomic levels are currently included in the database,
various datasets, showcasing its utility in characterizing along with the corresponding phylogenies and sequence
dysregulated pathways and improving metabolite alignments. Despite the growth in genomic data,
identification. DTINet, a statistical pipeline, uses the quality of functional annotations and orthology
36
multimodal network integration to predict drug-target assignments is still good at 80% coverage. With
interactions, boosting the accuracy of predictions enhanced online services and application programming
and uncovering novel drug-cyclooxygenase protein interface searches, users may investigate evolutionary
interactions. This highlights potential implications in histories and functional annotations. 44
inflammation disease prevention. 37 The Orthologous Matrix database has been updated
An additional study discusses artificial intelligence’s with new species and improved tools for orthology
transformative role in drug discovery, formulation, analysis. New features include Ancestral Genome
and pharmaceutical dosage form testing. It highlights pages and a Local Synteny Viewer for genomic
artificial intelligence’s ability to analyze biological data comparisons. The paper discusses enhancements in
for targeted drug discovery. Artificial intelligence can search functionality and Gene Ontology annotations
optimize research processes, reduce development costs, for Hierarchical Orthologous Groups. The Orthologous
and enhance drug candidate evaluation. Personalized Matrix database is accessible online, providing
medicine is facilitated through artificial intelligence, resources for studying gene families and evolutionary
improving treatment outcomes and patient adherence. history. OrthoDB is a comprehensive resource for
45
The review emphasizes the potential of artificial evolutionary and functional annotations of orthologs,
intelligence in enhancing drug development and patient covering a vast number of organisms, including
care. The study examines how artificial intelligence may eukaryotes, prokaryotes, and viruses, with plans to
38
be used in drug development, emphasizing how it can significantly increase bacterial sampling. The user
be used to anticipate protein structures and interactions interface has been enhanced to improve usability,
between drugs and targets while tackling issues such offering three views: A list of orthologous groups, a
as data quality and technology limitations. The study detailed view of these groups, and a gene-centric view.
39
discusses the critical issue of rising sea levels, which are OrthoDB allows users to upload their data for analysis
projected to increase by 20 cm by 2050. It highlights the and provides evolutionary annotations, such as phyletic
potential displacement of up to 1.2 billion people due to profiles and evolutionary rates, which are unique to the
this environmental threat. A high-level United Nations resource. The resource is publicly accessible, facilitating
meeting was convened to address the existential threat comparative studies and metagenomics. Another study
46
posed by rising sea levels. Heavy metal concentrations discusses the significance of QSAR modeling in drug
40
varied widely, with the highest levels observed in iron, discovery, emphasizing its efficiency in identifying lead
manganese, and zinc. The research highlights significant candidates. It highlights the need for advanced machine
spatial variability in contamination levels influenced by learning algorithms to manage large datasets in QSAR
traffic and anthropogenic activities. 41 applications. The authors note that successful QSAR
Another study employs machine learning to predict projects require interdisciplinary collaboration and
water quality, especially total coliform presence, critical thinking from scientists. The paper also addresses
using an Indian dataset. Gradient boosting regression common pitfalls in QSAR modeling, including a lack
produces good accuracy, with conductivity and of understanding of best practices. Overall, it provides
temperature playing critical roles. A website called recommendations for improving QSAR-based virtual
42
OrthoVenn2 allows users to compare whole-genome screening methodologies in drug discovery. 47
orthologous clusters from up to 12 different species. Edge computing refers to the practice of processing
A Venn diagram and an interactive occurrence pattern data closer to the data source or “edge” of the network,

Volume 22 Issue 3 (2025) 93 doi: 10.36922/AJWEP025070041