Mapping the functional gene-scape of the oceans under conditions of global change

背景：本提案建立在我们的工作基础上，该工作通过Mocks实验室（Duncan, 2020）的大规模宏基因组测序构建宏基因组组装基因组（MAGs），确定了极性和非极性微生物在基因组水平上的功能差异。在这些数据中，我们最近发现了与叶绿素浓度卫星观测相关的光合代谢途径中的不同成分，这表明基因功能与海洋区域之间存在直接联系。目标1：学生将对广泛的公共组学数据集进行功能性注释，包括（Sunagawa, 2015）和（Zhang, 2020）中的数据集，以及正在进行的极地MOSAiC巡航的数据集。所有数据都将使用EBI MGnify管道或类似工具以与Tara Oceans数据相当的格式进行注释。通过这种方式，学生将获得使用生物信息学工具和处理大型测序数据集的经验。然后，她将对在北极和南极群落中发现的基因功能进行生物信息学比较，并将其与已经从相同数据中获得的mag基因功能进行比较。目标2：学生将在注释数据上应用非负矩阵分解和相关的机器学习方法来识别表征表面海洋区域之间变化的共发生函数组。NMF已被用于研究海洋环境中宏基因组数据的功能模式（Jiang, 2012）。为了研究分组背后的生物和代谢过程，学生将应用聚类和可视化方法，如网络表示，来识别组内表征其变化的关键功能。然后将探讨群体与包括卫星数据在内的环境元数据之间的联系。这将提供将无监督机器学习方法应用于混合数据集的经验，并与专家合作，从结果中识别环境见解。目标3：基于降维模型和环境元数据之间的关联，学生将开发从环境条件预测基因功能的模型。模型已经成功地用于预测浮游植物的分类结构（Bracher, 2017）以及以分类为中间步骤的群落基因功能（Larsen, 2015）。我们的目标是在不参考分类学的情况下预测基因功能。学生将使用贝叶斯网络来解决这个问题，特别是从宏基因组数据中学习网络结构。这将使他们能够从独立的环境变量中推断基因功能。安全性与挑战：由于该项目建立在所有3个目标的坚实基础之上，因此几乎没有证据表明这项工作不会成功。然而，挑战在于发展生物信息学和机器学习方面的技能，以确定对环境有意义的结果。通过获得的技能集，有可能通过在高度综合和多学科的环境中工作来确定协同作用。

Context: This proposal builds on our work which identified differences in function between polar and non-polar microbes at the genomic level, through constructing metagenome assembled genomes (MAGs) from large-scale metagenomic sequencing from Mocks lab (Duncan, 2020). Within this data, we have recently identified varying components within photosynthetic metabolic pathways correlated with satellite observations of chlorophyll concentration, suggesting a direct link between gene function and ocean regions. Objective 1: The student will functionally annotate extensive public omics datasets including those in (Sunagawa, 2015) and (Zhang, 2020) as well as from the ongoing polar MOSAiC cruise when it becomes available. All data will be annotated in a format comparable to Tara Oceans data using the EBI MGnify pipeline or similar tools. In this way the student will gain experience with bioinformatics tools, and handling large sequencing datasets. She will then perform a bioinformatics comparison of gene functions identified in the Arctic and Antarctic communities, as well as comparing this to the gene functions of MAGs already available from the same data.Objective 2: The student will apply non-negative matrix factorisation and related machine learning approaches on the annotated data to identify groups of cooccurring functions characterising variation between surface ocean regions. NMF has been used to investigate functional patterns in metagenomics data in marine contexts (Jiang, 2012). To investigate the biological and metabolic processes underlying the groupings, the student will apply clustering and visualisation methods like network representations to identify key functions within groups which characterise their variation. Associations between groups and environmental metadata including satellite data will then be explored. This will provide experience in applying unsupervised machine learning approaches to mixed data sets, and working with experts to identify environmental insights from the results.Objective 3: Based on associations identified between the reduced dimension models and environmental metadata, the student will develop models to predict gene functions from environmental conditions. Models have been successfully used to predict taxonomic structure (Bracher, 2017) as well as community gene function in phytoplankton using taxonomy as an intermediate step (Larsen, 2015). We aim to predict gene function without reference to taxonomy. The student will approach this using Bayesian networks, in particular learning the network structure from the metagenomic data. This will enable them to infer gene functions from the independent environmental variables. Security vs challenge: As this project builds on strong foundations for all 3 objectives, there is little evidence that this work will not be successful. Yet, the challenge is to develop skills in bioinformatics and machine learning to identify environmentally meaningful results. With the acquired skill set, there is potential to identify synergies through working in a highly integrative and multidisciplinary environment.