|
Development of an AI-Based Predictive Model for Assessing Genetic Vulnerability to Marine Toxins in Seafarers and Divers
|
Salimeh Khanamani Falahati-pour   , Amirreza Ghasemloo , Somayyeh Karami-Mohajeri , Zohreh Oghabian , Soudeh Khanamani Falahati-pour , Motahareh Soltani * |
| Pistachio Health Research Center, Rafsanjan University of Medical Sciences, Rafsanjan, Iran , Motahareh.soltani@gmail.com |
|
|
Abstract: (31 Views) |
Background and Aim: Marine toxins pose a significant threat to the health of seafarers and divers, particularly in key regions such as the Persian Gulf and the Gulf of Oman. Individual susceptibility to these toxins is often associated with genetic variations, specifically single nucleotide polymorphisms (SNPs) in genes involved in toxin metabolism. This article presents a conceptual artificial intelligence (AI)-based framework for predicting this genetic vulnerability. The primary innovation lies in the multi-source integration of genetic data, environmental data (from sensors and satellites), and toxin knowledge bases (e.g., ToxCast), with connectivity to telemedicine and real-time alert systems.
Methods: The proposed framework is built upon three core components: 1) individual genetic data (SNP profiles in toxin metabolism-related genes), 2) environmental data collected from marine sensors and satellite imagery (including concentrations of algal toxins such as ciguatoxin, saxitoxin, and domoic acid, water quality, and temperature), and 3) toxin knowledge bases such as ToxCast and T3DB. Three candidate algorithmic categories -deep neural networks, random forests, and support vector machines (SVM)- are considered for risk modeling. The model evaluation protocol follows training/validation/test data splitting with cross-validation, and performance metrics including accuracy, precision, recall, and F1-score (with AUC reported if applicable) are defined. Final model selection will be based on a comparison of these metrics. The geographical scope for field data collection encompasses shipping routes and diving operations in the Persian Gulf and the Gulf of Oman.
Results: Given the conceptual and framework-oriented nature of this study, no numerical or experimental results are reported at this stage. The main outputs include: a) model architecture design and data processing flowchart, b) specifications of data sources and collection methods (including marine sensors, satellite imagery, toxicology databases, and genetic samples), c) a quantitative evaluation protocol with defined metrics, and d) a clear delineation of the research innovation (integration of genetics, environmental data, and AI) and its applications (in diving/occupational medicine, maritime emergency response via telemedicine, harmful algal bloom monitoring, and early warning systems). Key implementation challenges -such as computational infrastructure in maritime environments, data quality and completeness, and ethical considerations related to genetic data- and their implications for system design are also discussed.
Conclusion: This framework provides a clear pathway for implementing and evaluating a personalized risk prediction system, demonstrating its potential to enhance occupational health, reduce high-risk exposures, and support remote clinical decision-making. Empirical validation with real-world data and reporting of quantitative performance metrics represent the next phase of this research.
|
|
| Keywords: Genetic predisposition to disease, Marine toxins, Artificial intelligence, Seafarers, Health prediction |
|
|
Full-Text [PDF 533 kb]
(35 Downloads)
|
Type of Study: Original Article |
Subject:
Marine Medicine
Received: 2025/08/8 | Accepted: 2025/11/19 | Published: 2025/12/1
|
|
|
|
|
|
|
| Send email to the article author |
|
|
|
| Add your comments about this article |
|
|
|
