Integrative predictive modeling of VHSV transmission in olive flounder aquaculture using infection dynamics and environmental surveillance

다국어 초록 (Multilingual Abstract)

Viral hemorrhagic septicemia virus (VHSV) is one of the most significant viral pathogens causing sustained economic losses in olive flounder (Paralichthys olivaceus) aquaculture. The infection and transmission of VHSV are governed by complex interactions among host susceptibility, water temperature, rearing conditions, and hydrodynamic connectivity. With recent advances in environmental DNA/RNA (eDNA/eRNA)-based disease surveillance, machine learning–based risk prediction, and hydrodynamic dispersion modeling, there is an increasing need to shift VHSV management from a diagnosis-oriented approach toward a predictive and management-oriented framework. The objective of this study was to elucidate the transmission mechanisms of VHSV in olive flounder aquaculture by integrating experimental infection dynamics, field-based eRNA monitoring, machine learning–based risk prediction, and hydrodynamic spread simulation, and to establish an integrated framework for proactive disease management. Controlled infection trials were conducted using olive flounder with body weights of 10, 100, and 200 g under water temperature conditions of 8, 13, and 18 oC. Additionally, outlet-water–based eRNA monitoring was performed at commercial olive flounder farms in Jeju Island. Experimental and field-derived datasets were integrated to develop a machine learning–based VHSV risk prediction model using water temperature, body weight, outlet-water detection status, and clinical signs as input variables. Furthermore, temperature-dependent virus dispersal scenarios were simulated using a hydrodynamic particle-tracking model incorporating regional ocean current circulation. The results demonstrated that VHSV susceptibility in olive flounder was clearly differentiated according to body weight and water temperature. Field-based susceptibility thresholds were identified, indicating a substantial reduction in infection risk at body weights above approximately 158 g and water temperatures exceeding 18.7 oC. eRNA-based monitoring was shown to be applicable as a non-invasive tool for detecting infection occurrence at aquaculture sites. The minimum infectious dose (MID) of VHSV was determined to be ≥10³ copies mL⁻¹, and smaller fish exhibited relatively higher virus shedding potential, suggesting their role as contributors to enhanced within-population transmission. Among the evaluated machine learning–based VHSV risk prediction models, XGBoost exhibited the highest predictive performance, achieving an accuracy of 85% and an area under the curve (AUC) of 0.85. Hydrodynamic spread simulation further revealed that viral persistence and dispersal range increased markedly under low-temperature conditions, with transmission patterns varying according to regional hydrodynamic characteristics. In conclusion, this study provides an integrated and quantitative understanding of VHSV transmission dynamics in olive flounder aquaculture by combining experimental, field-based, and modeling approaches. The proposed framework supports infection epidemiology–based decision-making, early warning, and spatial risk assessment, thereby facilitating a transition from reactive disease control to predictive and proactive management. These findings offer a scientific basis for improving VHSV management strategies and enhancing biosecurity in olive flounder aquaculture systems.

국문 초록 (Abstract)

바이러스성출혈성패혈증 바이러스(VHSV)는 넙치(Paralichthys olivaceus) 양식에서 지속적인 피해를 유발하는 대표적 바이러스성질병으로, 그 감염과 전파는 숙주 감수성, 수온, 사육 환경, 해류 연결성 등 다양한 요인의 상호작용에 의해 결정된다. 최근 eDNA/RNA 기반 질병 모니터링, 머신러닝을 활용한 질병 예측 모델, 그리고 해류 유동 기반 질병 확산 시뮬레이션 기술이 빠르게 발전함에 따라, VHSV 관리 전략을 기존의 진단 중심 접근에서 예측·관리 중심으로 전환할 필요성이 제기되고 있다. 본 연구의 목적은 실험 감염역학, 현장 eRNA 모니터링, 머신러닝 기반 VHSV위험 예측, 그리고 해류 기반 확산 분석을 통합하여 넙치 양식장에서의 VHSV전파 메커니즘을 규명하고, 선제적 질병관리를 위한 통합 프레임워크를 구축하는 것이다. 이를 위해 8, 13, 18 °C의 수온 조건에서 체중 10, 100, 200 g 넙치를 대상으로 감염시험을 수행하였으며, 제주 지역 양식장을 대상으로 배출수 기반 eRNA 모니터링을 실시하였다. 또한 실험 및 현장 데이터를 통합하여 수온, 체중, 배출수 검출 여부, 임상증상을 입력 변수로 하는 머신러닝 기반 VHSV 위험 예측 모델을 개발하였고, 해류 유동을 고려한 입자추적 모델을 통해 수온별 바이러스 확산 시나리오를 시뮬레이션하였다. 그 결과, 넙치의 VHSV 감수성은 체중과 수온에 따라 뚜렷하게 구분되었으며, 체중 약 158 g 이상 및 수온 18.7 °C 이상에서 감염 위험이 현저히 감소하는 현장 기반 감수성 임계값을 도출하였다. 또한 eRNA 기반 모니터링이 양식장에서의 감염 발생을 비침습적으로 탐지할 수 있는 가능성을 확인하였다. 넙치의 최소감염농도(MID)는 103 copies/mL이상으로 확인되었으며, 소형 개체는 상대적으로 높은 바이러스 배출 잠재력을 보여 집단 내 전파 증폭자로 작용할 가능성이 높았다. 머신러닝 모델 중 XGBoost는 정확도 85%와 AUC 0.85로 가장 우수한 예측 성능을 보였으며, 해류 기반 확산 분석에서는 저수온 조건에서 바이러스의 잔존 및 확산 범위가 크게 증가하며, 지역별 유동 특성에 따라 전파 양상이 달라짐을 확인하였다. 종합하면, 본 연구는 실험·현장·모델링 결과를 통합하여 VHSV 전파를 정량적으로 설명할 수 있는 근거를 제시하였다. 이를 통해 감염역학 기반 의사결정, 조기경보, 공간 기반 위험지역 설정을 가능하게 함으로써, 기존의 사후대응형 관리에서 예측 기반의 선제적 관리로의 전환에 기여할 것으로 기대된다. 또한 본 연구는 넙치 양식장에서의 VHSV 전파 이해를 심화시키고, 현장 적용이 가능한 통합 관리 전략 수립에 기초 자료를 제공할 것이다.

목차 (Table of Contents)

• General introduction 1
• Chapter I: VHSV infection dynamics in olive flounder under host and environmental variability 5
• 1. Introduction 7
• 2. Materials and methods 9
• 2.1. Shedding dynamics of VHSV in relation to water temperature and fish weight 9
• 2.1.1. Virus strain and culture 9
• 2.1.2. Fish and sampling 10
• 2.1.3. Iron flocculation 12
• 2.1.4. RNA extraction and qPCR 14
• 2.1.5. PCR efficiency and limit of detection 15
• 2.1.6. Statistical analysis 18
• 2.2. Integrating viral decay and minimum infectious dose to understand VHSV transmission 19
• 2.2.1. Virus strain and culture 19
• 2.2.2. Fish and sampling 20
• 2.2.3. RNA extraction and qPCR 21
• 2.2.4. Stability of VHSV 22
• 2.2.5. Statistical analysis 24
• 2.3. Influence of stocking density on VHSV pathogenicity 25
• 2.3.1. Virus strain and culture 25
• 2.3.2. Fish and sampling 26
• 2.3.3. RNA extraction and qPCR 27
• 2.3.4. Statistical analysis 27
• 3. Results 28
• 3.1. Shedding dynamics of VHSV in relation to water temperature and fish weight 28
• 3.1.1. qPCR performance (efficiency and sensitivity) 28
• 3.1.2. Mortality and survival 30
• 3.1.3. Effects of water temperature and fish weight on VHSV pathogenicity in fish 34
• 3.1.4. Effects of water temperature and fish weight on VHSV shedding in rearing water 38
• 3.1.5. Statistical modeling of factors associated with viral shedding 40
• 3.2. Integrating viral decay and minimum infectious dose to understand VHSV transmission 44
• 3.2.1. Mortality and survival 44
• 3.2.2. Determination of the minimum infectious dose (MID) of VHSV 46
• 3.2.3. Statistical modeling of infection probability by viral dose and fish weight 49
• 3.2.4. Viral decay kinetics of VHSV in seawater across environmental factors 51
• 3.3. Influence of stocking density on VHSV pathogenicity 53
• 3.3.1. Pathogenicity of VHSV under different stocking densities 53
• 3.3.2. Statistical modeling of infection probability by stocking density 56
• 4. Discussion 58
• 5. Conclusion 66
• Chapter II: Field-based environmental RNA (eRNA) surveillance 67
• 1. Introduction 68
• 2. Materials and methods 72
• 2.1. Development of plankton-net concentration for efficient monitoring of VHSV 72
• 2.1.1. Virus strain and culture 72
• 2.1.2. Plankton and Fish 73
• 2.1.3. Plankton-virus co-culture experiments 73
• 2.1.4. Assessment of viral adsorption in fish fecal samples 75
• 2.1.5. Assessing viral concentration with a plankton net via in vivo challenge 76
• 2.1.6. Iron flocculation 77
• 2.1.7. RNA extraction and qPCR 78
• 2.1.8. Statistical analysis 78
• 2.2. Environmental RNA-based surveillance of VHSV in olive flounder aquaculture: Detection dynamics and risk assessment 80
• 2.2.1. Laboratory assessment of VHSV infection in fish and rearing water 80
• 2.2.2. Iron flocculation 81
• 2.2.3. RNA extraction and qPCR 82
• 2.2.4. VHSV monitoring in aquaculture farms 83
• 2.2.5. Statistical analysis 84
• 3. Results 85
• 3.1. Development of plankton-net concentration for efficient monitoring of VHSV 85
• 3.1.1. Assessing qPCR inhibition in VHSV quantification in plankton and fecal samples 85
• 3.1.2. Assessment of VHSV-plankton interaction via co-culture experiments 87
• 3.1.3. Assessing plankton nets for VHSV concentration in rearing water with infected fish 90
• 3.2. Environmental RNA-based surveillance of VHSV in olive flounder aquaculture: Detection dynamics and risk assessment 94
• 3.2.1. qPCR performance (efficiency and sensitivity) 94
• 3.2.2. Assessment of VHSV shedding dynamics in rearing water from infected fish 95
• 3.2.2.1. Quantification of VHSV shed by artificially infected fish in rearing water 95
• 3.2.2.2. ROC analysis for determining VHSV shedding-risk thresholds 96
• 3.2.3. VHSV monitoring in aquaculture farms 100
• 3.2.3.1. VHSV detection in fish spleen and outlet water samples and their associations 100
• 3.2.3.2. Relationship between VHSV infection and water temperature or fish weight 101
• 4. Discussion 108
• 5. Conclusion 114
• Chapter III: Developing predictive models for VHSV transmission: Machine learning models and hydrodynamic particle-tracking simulations 115
• 1. Introduction 116
• 2. Materials and methods 121
• 2.1. Development of machine-learning models for infection-risk prediction 121
• 2.1.1. Sample data collection 121
• 2.1.2. Data preprocessing 122
• 2.1.3. Model development for predicting VHSV infection risk 123
• 2.1.3.1. Multinomial logistic regression 126
• 2.1.3.2. Decision tree 126
• 2.1.3.3. Random forest 126
• 2.1.3.4. XGBoost 127
• 2.1.3.5. Multilayer Perceptron (MLP) 127
• 2.2. Hydrodynamic particle-tracking modeling of VHSV spread scenarios 128
• 2.2.1. Site selection and Lagrangian TRANSport modeling 128
• 2.2.2. Simulation of the transmission range of infectious VHSV under different water temperatures 129
• 3. Results 132
• 3.1. Development of machine-learning models for infection-risk prediction 132
• 3.1.1. Model-specific performance 132
• 3.1.1.1. Multinomial logistic regression 134
• 3.1.1.2. Decision tree 135
• 3.1.1.3. Random forest 137
• 3.1.1.4. XGBoost 139
• 3.1.1.5. Multilayer Perceptron (MLP) 141
• 3.1.2. Optimal model selection 143
• 3.2. Hydrodynamic particle tracking modeling of VHSV spread scenarios 147
• 3.2.1. Temperature-dependent decay modeling and simulation of VHSV persistence 147
• 3.2.2. Evaluating the potential risk of disease transmission in each site under different water temperatures 152
• 3.2.2.1. Hallim 152
• 3.2.2.2. Daejeong 155
• 3.2.2.3. Namwon 158
• 3.2.2.4. Seongsan 161
• 3.2.2.5. Gujwa 164
• 3.2.2.6. Sinji 167
• 3.2.2.7. Summary of temperature-dependent VHSV spread simulation across different zones 170
• 4. Discussion 172
• 5. Conclusion 178
• Chapter IV: Integrated strategies for controlling VHSV in olive flounder aquaculture 179
• 1. Introduction 180
• 2. Integrated framework for VHSV control 182
• 2.1. Farm-level VHSV management strategies 184
• 2.1.1. Adjustment of stocking time and fish weight 184
• 2.1.2. Stocking registration and traceability system 185
• 2.1.3. Disinfection of inlet and outlet water 186
• 2.1.4. Biosecurity management in feed, environment, and facility hygiene 188
• 2.1.5. Regional considerations for adaptive control 190
• 2.2. Non-invasive outlet-water eRNA surveillance 192
• 2.3. Predictive modeling and spatial risk management 194
• 2.3.1. Integration of machine learning–based prediction and hydrodynamic simulation 194
• 2.3.2. Analysis of viral dispersal patterns in representative zones 196
• 2.3.3. Region-specific VHSV risk management strategies based on model outputs 203
• 3. Conclusion 205
• Acknowledgment 206
• References 208