RII Track-2 FEC: Highly Predictive, Explanatory Models to Harness the Life Science Data Revolution

数据规模和可用性的戏剧性增长正在深刻地重塑生命科学。数据获取和可用性正在超过我们的分析能力，包括开发代表我们对生物过程的知识的模型。这个由怀俄明州、蒙大拿州和内华达州的三所大学合作的项目将通过我们联盟领导的研究和教育努力，满足生命科学领域的这一迫切需求。一些类型的模型可以很好地拟合观测数据，但缺乏通用性，无法外推到新的设置或未来的时间点。相反，其他类型的模型可以更通用，但对单个数据集的适合性较差。在我们的研究中，我们将发展这些权衡的知识和结合不同类型模型的优势特征的方法。每年，我们的联盟将培训12名博士后研究人员组成的不同队列，学习尖端建模技术，并为他们的工作做好准备。我们三个机构的项目调查人员和博士后研究人员将创建一个由数据科学家组成的综合、高度协作和跨学科的联盟。我们将开发教育工具，帮助传播我们开发的方法，促进高维数据在生命科学中的有效利用。数据规模和可用性的急剧增加正在深刻地重塑生命科学的所有领域。来自DNA测序仪、环境传感器、并行全球研究和图像(以及其他许多)的数据获取和可用性，跨越时间和空间，正在超过我们的分析能力，包括开发代表我们对生物过程的知识的模型。我们将通过我们在怀俄明大学、蒙大拿大学和内华达-里诺大学的联盟领导的研究和教育努力，解决生命科学中的这一差距。我们将竞争并进一步开发用于多维数据的计算、统计和机器学习方法，以开发用于生命科学的高度预测性和解释性模型。我们将测试和改进方法，开发关键工具，以驾驭数据革命。我们将把它们应用于生命科学中的三个跨尺度领域，并推进我们对关键生态和进化现象的机械理解。通过将目前使用来自多个学科的现有技术的科学家联盟聚集在一起，包括计算机科学和应用数学，并用模拟和真实数据竞争这些技术，我们将评估现有方法的有效性。此外，我们将在这些方法的基础上开发新的混合建模技术，并扩大线性和非线性稀疏模型的使用，以最大限度地提高应用于高维数据集的模型的预测精度和机理洞察力。我们将把这些技术应用于生命科学中的关键挑战，包括将表型映射到基因组数据，在高度多样化的系统中对群落动态进行建模，以及解开水生态系统中不同时间尺度驱动因素之间的相互作用。每年，我们的联盟都将培训12名博士后研究人员的建模技术，并为他们的工作做好准备。怀俄明州、蒙大拿州和内华达州的项目调查人员和博士后研究人员将创建一个综合的、高度协作的跨学科数据科学家联盟。我们将开发教育工具来帮助传播我们开发的方法，促进生命科学中高维数据的有效利用。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Dramatic increases in the scale and availability of data are profoundly reshaping the life sciences. Data acquisition and availability are outpacing our capacity for analysis, including the development of models that represent our knowledge of biological processes. This collaborative project among three universities in Wyoming, Montana, and Nevada will address this pressing need in the life sciences through research and education efforts led by our consortium. Some types of models can fit observed data very well, but lack generality and the ability to extrapolate to novel settings or future time points. Conversely, other types of models can be more general, but provide a poorer fit to individual data sets. In our research we will develop knowledge of these trade-offs and methods that combine advantageous features of different types of models. In each year our consortium will train a diverse cohort of twelve postdoctoral researchers in cutting edge modeling techniques and prepare them for the workforce. The project investigators and postdoctoral researchers at our three institutions will create an integrated, highly collaborative and interdisciplinary consortium of data scientists. We will develop educational tools to aid the dissemination of the methodologies we develop, promoting the efficient use of high dimensional data in the life sciences.Dramatic increases in the scale and availability of data are profoundly reshaping all domains in the life sciences. Data acquisition and availability from DNA sequencers, environmental sensors, parallel global studies, and imagery (among many others), across time and space, are outpacing our capacity for analysis, including the development of models that represent our knowledge of biological processes. We will address this gap in the life sciences through research and education efforts led by our consortium at the University of Wyoming, the University of Montana, and the University of Nevada–Reno. We will compete and further develop computational, statistical, and machine learning methods for multi-dimensional data to develop highly predictive and explanatory models for the life sciences. We will test and refine methods and develop critical tools for harnessing the data revolution. We will apply them to three cross-scale domains within the life sciences and advance our mechanistic understanding of key ecological and evolutionary phenomena. By bringing together a consortium of scientists currently using existing techniques from multiple disciplines, including computer science and applied mathematics, and competing these techniques with simulated and real data, we will evaluate the efficacy of existing methods. Further, we will build on these methods to develop novel hybrid modeling techniques and expanded use of linear and non-linear sparse models to maximize both the predictive accuracy and mechanistic insight of models applied to high dimensional data sets. We will apply these techniques to critical challenges in the life sciences, including mapping phenotypes to genomic data, modeling community dynamics in highly diverse systems, and disentangling the interplay among different temporal scales of drivers in aquatic ecosystems. In each year our consortium will train a diverse cohort of twelve postdoctoral researchers in modeling techniques and prepare them for the workforce. The project investigators and postdoctoral researchers in Wyoming, Montana, and Nevada will create an integrated, highly collaborative and interdisciplinary consortium of data scientists. We will develop educational tools to aid the dissemination of the methodologies we develop, promoting the efficient use of high dimensional data in the life sciences.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.