摘要

      海水覆盖了地球表面70%以上，含有溶解有机物（DOM）和微生物。海洋是大气中有机物和遗传物质的重要来源之一，从而影响气候变化和全球变暖。因此，关于海水和大气中有机物和遗传物质的特征和起源的大数据使我们能够长期了解生态系统和气候的变化。此外，深度学习技术的应用有望预测包括海水在内的环境中污染物的发生。因此，在这项研究中，我的目的是1）分析关于海水中有机物特征和起源的大数据，2）调查关于海水中遗传物质特征和来源的大数据；3）开发深度学习技术，预测海水中遗传物的发生。首先，为了确定海水中DOM的纬度特征，沿着从西太平洋（36°N）到南大洋（74°S）的横断面收集海水。根据海水中DOM的光学特性，DOM似乎受到中纬度陆地来源的影响，而微生物在低纬度产生，在高纬度在海洋中产生。Orbitrap MS在分子水平上对DOM的表征显示出与光学性质相似的结果，尽管在整个站点中主要是木质素样成分。此外，利用Orbitrap MS分析了气溶胶中的DOM特征，表明木质素类成分主要影响从海洋到大气的过程，这可能是由于木质素的疏水性。另一方面，在存在海冰的南大洋罗斯海（RS）收集的气溶胶中，类脂成分占主导地位。RS气溶胶中的这些DOM特征与海冰的DOM特征有关，这表明通过风吹效应从海冰表面向大气中雾化的可能性。因此，为了应对气候变化，第一项研究在全球范围内描述了海水和海冰中DOM特征对大气的可能影响。其次，在从西太平洋（36°N）到南大洋（74°S）的横断面上，研究了韩国破冰船R/V Araon号（巡航总距离：14942公里）上抗生素抗性基因（ARGs）的丰度和多样性以及细菌群落的组成。在西太平洋，ARGs的总绝对丰度为3.0×106±1.6×106拷贝/mL，在塔斯曼海（37°S）的一个站记录到最高值（7.8×106拷贝/mL）。南大洋ARGs的总绝对丰度比西太平洋低1.8倍，在向南极洲的Terra Nova湾移动时略有增加（0.7倍），这可能是由于自然陆地来源或人类活动造成的。β-内酰胺和四环素抗性基因在所有样本中占主导地位（88–99%），表明它们可能是影响ARGs总丰度的关键成分。相关性和网络分析表明，拟杆菌门、蓝细菌门、Margulisbacteria和变形菌门与ARGs呈正相关，表明这些细菌可能携带ARGs。第二项研究强调了公海系统中ARG分布的纬度剖面，以及在全球范围内监测这种新出现的污染物的新见解。第三，使用定量实时聚合酶链式反应和下一代测序对2018年6月和2019年5月从韩国休闲海滩每1-5小时采集的海水样本进行分析。我们发现，降雨量和潮汐水平对总ARGs的相对丰度有很大影响，增加了1.9×103倍。特别是，ARGs的升高水平在降雨后维持了长达32小时。与污水相关的ARG丰度的增加表明，下水道综合溢流（CSO）可能是导致ARG数量和多样性增加的主要因素。此外，我们使用传统的长短期记忆（LSTM）、LSTM卷积神经网络（CNN）混合模型和输入注意力（IA）-LSTM，提出了主要在降雨后海岸发现的ARGs发生的预测模型。当单独预测ARGs的发生时，传统的LSTM和IA-LSTM在训练和测试期间表现出较差的R2值。相比之下，LSTM-CNN在准确性上比传统LSTM和IA-LSTM提高了2-6倍。然而，当同时预测所有ARG的发生时，与LSTM-CNN相比，IA-LSTM模型总体表现出优越的性能。此外，使用IA-LSTM模型研究了环境变量对预测的影响，并确定了影响每个ARG的输入变量的范围。因此，第三项研究证明了根据各种环境变量预测海滩上主要ARG发生和分布的可能性，其结果有望有助于管理休闲海滩上的ARG发生。

Abstract

Seawater covers more than 70 percent of surface of the Earth and contains dissolved organic matters (DOM) and microorganisms. Sea is one of the important sources of organic and genetic matters of the atmosphere, which can consequently affect climate change and global warming. Accordingly, big data on the haracteristics and origins of organic and genetic matters in seawaters and atmosphere allows us to understand the changes in ecosystem and climate in the long term. Moreover, the application of deep learning technique is expected o be promising in predicting the occurrence of pollutants in the environment including seawaters. Therefore, in this study, I aimed 1) to analyze big data on the characteristics and origins of organic matter in seawaters, 2) to investigate big data on the characteristics and origins of genetic matter in seawaters, and 3) to develop deep learning techniques and predict the occurrence of the genetic matters in seawaters. First, to identify latitudinal characteristics of DOM in seawater, seawaters were collected along a transect from the western Pacific Ocean (36°N) to the Southern Ocean (74°S). Based on optical properties of DOM in seawaters, it seems that DOM were affected by terrestrial sources in the mid-latitude, while microbiologically produced in the low-latitude and produced within the ocean in the high-latitude. DOM characterization at molecular-level by Orbitrap-MS showed similar result to optical properties despite dominant lignin-like component in overall stations. In addition, the DOM characteristics in aerosol were analyzed with Orbitrap-MS, indicating that lignin-like component affects dominantly from the ocean to the atmosphere probably due to hydrophobicity of lignin. On the other hand, lipid-like component was dominant in aerosol collected from Ross Sea (RS) in the Southern Ocean where sea ice exists. These DOM characteristics in aerosol of RS were related to those of sea ice, suggesting the possibility of aerosolization from the surface of sea ice to the atmosphere by wind-blown effect. Therefore, in response to climate change, the first study described the possible effects of DOM characteristics in seawater and sea ice on the atmosphere on a global scale. Second, the abundance and diversity of Antibiotic resistance genes (ARGs) and the composition of bacterial communities were investigated along a transect from the western Pacific Ocean (36°N) to the Southern Ocean (74°S) onboard the Korean icebreaker R/V Araon (total distance of the cruise: 14,942 km). The total absolute abundance of ARGs was 3.0×106 ± 1.6×106 copies/mL in the western Pacific Ocean, and the highest value (7.8×106 copies/mL) was recorded at a station in the Tasman Sea (37°S). The total absolute abundance of ARGs in the Southern Ocean was 1.8-fold lower than that in the western Pacific Ocean, and slightly increased (0.7–fold) as they headed toward Terra Nova Bay in Antarctica, probably resulting from natural terrestrial sources or human activity. β-Lactam and tetracycline resistance genes were dominant for all samples (88–99%), indicating that they may the key components affecting the total abundance of ARGs. Correlation and network analyses showed that Bdellovibrionota, Bacteroidetes, Cyanobacteria, Margulisbacteria, and Proteobacteria were positively correlated to ARGs, suggesting that these bacteria potentially carry ARGs. The second study highlights the latitudinal profile of ARG distribution in open-ocean system and new insight to monitor this emerging pollutant on a global scale. Third, seawater samples, collected every 1‒5 hours in June 2018 and May 2019 from recreational beach in Korea, were analyzed using quantitative real-time polymerase chain reaction and next-generation sequencing. We found a substantial influence of rainfall and tidal levels on the relative abundance of total ARGs showed 1.9 × 103 fold increases. In particular, the elevated levels of ARGs were maintained for up to 32 hours after rainfall. An increased abundance of sewage‐related ARGs suggested that combined sewer overflow (CSO) may be the major factor contributing to the increase in the number and diversity of ARGs. Moreover, we proposed prediction models for ARGs occurrence that are primarily found on the coast after rainfall using a conventional long short-term memory (LSTM), LSTM-convolutional neural network (CNN) hybrid model, and input attention (IA)-LSTM. When individually predicting ARGs occurrence, the conventional LSTM and IA-LSTM exhibited poor R2 values during training and testing. In contrast, the LSTM-CNN exhibited a 2–6-times improvement in accuracy over those of the conventional LSTM and IA-LSTM. However, when predicting all ARGs occurrence simultaneously, the IA-LSTM model exhibited a superior performance overall compared to that of LSTM-CNN. Additionally, the influence of environmental variables on prediction was investigated using the IA-LSTM model, and the ranges of input variables that affect each ARG were identified. Consequently, the third study demonstrated the possibility of predicting the occurrence and distribution of major ARGs at the beach based on various environmental variables, and the results are expected to contribute to management of ARG occurrence at a recreational beach.

https://scholarworks.unist.ac.kr/handle/201301/57425