White Matter Connectivity and Network Analysis

在NICHD神经系统发育和可塑性部门以及儿科成像和组织科学项目(PPITS)的研究人员参与的一个项目中，NICHD使用了一个理论框架来预测髓鞘厚度的变化以及结节宽度的增加如何影响信号沿有髓轴突的传播。这两者都受到周围神经胶质细胞的调节，并依赖于轴突中存在的活动水平。理论预测在MATHEMICAL A中实现，并与实验观察值进行比较，最终目的是解决髓鞘胶质细胞在学习和可塑性中的作用。提交给神经元的手稿描述了关于星形胶质细胞在改变有髓轴突属性中的作用的详细实验结果。 我们还研究了髓鞘可塑性的理论方面以及这种适应性时间延迟的后果。特别是，我们研究了自适应时延对耦合振子系统稳定性的影响，这对大脑中振荡和同步的稳定性有一定的影响。在2014年发表在《神经科学》杂志上的一篇手稿中，我们展示了活动依赖性髓鞘形成的损害和适应性时间延迟的丢失如何导致神经元放电的高度和低同步性导致功能障碍(例如，阅读困难、精神分裂症、癫痫)的障碍。这表明，髓鞘的可塑性可能是维持发育中和成人大脑正常振荡活动所必需的。这项工作还将在2014年11月于华盛顿特区举行的神经科学学会年会上公布。 在与NIMH内壁研究计划中关键大脑动力学部分的研究人员一起进行的一个项目中，我们研究了加权复杂网络，特别是皮质和大脑网络。我们的研究揭示了新颖而稳健的权重组织在具有生物起源(神经、基因)的网络中尤其明显，但在不同的社会和语言(词)网络中也表现得尤为明显。此外，通过仿真，我们证明了这种网络结构可以使用局部学习规则来获得，所述局部学习规则基于节点之间过去的交互来调整网络中的权重。我们对这种学习规则进行了详细的仿真研究，这让人想起了时间延迟强化学习。我们表明，这种类型的学习增加了网络上所有路径的可达性，允许这样的复杂系统克服根深蒂固的结构/路径，并允许它适应新的挑战。这项工作已提交给IEEE网络科学与工程会刊。 我们还参与了两个与统计和机器学习有关的项目，但没有在生物医学统计学习项目(James Malley扮演Li)下报告。第一个涉及使用机器学习和线虫公开获得的大量遗传和其他数据，以调查从各种公共数据库提供的遗传、几何和其他信息可以在多大程度上估计线虫的已知神经连接性。我们表明，即使使用我们收集的一组有限的基因表达数据，也可以达到86%的准确率。我们预计，随着未来有更多的基因数据可用，准确性将会提高。这项工作于2013年11月在圣地亚哥举行的神经科学学会年会上发表。第二个项目涉及使用机器学习来估计小鼠的行为模式，以用于家庭笼养环境中啮齿动物的连续观察系统(SCORHE)。这项研究发表在2014年的《行为研究方法》杂志上。 作为与NICHD临床和发育基因组实验室的研究人员合作的项目，我们研究了通过对自闭症患者和正常人的皮肤细胞重新编程而创建的神经元培养。这是一项更大的研究的一部分，该研究旨在解决自闭症大脑在发育过程中发生的分子和细胞变化。在目前的研究中，目标是识别网络结构的变化，或不同细胞活动谱的变化，以区分正常细胞培养和自闭症患者的细胞培养。 在NICHD与儿科影像和组织科学计划(PPITS)的研究人员继续进行的项目中，我们对观察到的倾斜和重尾轴突直径分布进行了理论研究，并得出了优化信息上限(IUB)以及神经束的信息容量的参数形式。描述这项工作的手稿于2013年发表在《公共科学图书馆·综合》杂志上。在后续工作中，这种分布已经被实施并应用于模拟和实验数据，产生了改进的轴突直径分布的测量结果。

In a project with investigators from the Nervous System Development and Plasticity Section, NICHD, and the Program on Pediatric Imaging and Tissue Sciences (PPITS), NICHD we use a theoretical framework to predict how the changes in myelin thickness, as well as the increase in the nodal width, affects the propagation of the signals along a myelinated axon. Both of these are regulated by the surrounding glial cells and dependent on the level of activity present in an axon. The theoretical predictions are implemented in Mathematica and are compared with the experimentally observed values, with the ultimate goal of addressing the role of myelinating glia in learning and plasticity. Manuscript describing a detailed experimental findings on the role of astrocytes in modifying the myelinated axon properties is submitted to Neuron. We also study theoretical aspects of myelin plasticity and the consequences of such adaptive time delays. In particular, we studied the effect of adaptive time delays on the stability of the system of coupled oscillators, with implications to the stability of the oscillations and synchrony in the brain. In a manuscript published in Neuroscience, 2014, we showed how the impairment of activity-dependent myelination and the loss of adaptive time delays may contribute to disorders where hyper- and hypo-synchrony of neuronal firing leads to dysfunction (e.g., dyslexia, schizophrenia, epilepsy). This suggests that the myelin plasticity may be necessary to maintain normal oscillatory activity in the developing and adult brain. This work is also to be presented at the Annual Meeting of the Society for Neuroscience, Washington, D.C., in November of 2014. In a project with investigators of the Section on Critical Brain Dynamics in the Intramural Research Program at NIMH, we study weighted complex networks, in particular cortical and brain networks. Our study reveals novel and robust weight organization particularly pronounced in the networks with biological origin (neural, gene), but also in different social and language (word) networks. Additionally, using simulations, we show that such network architecture can be obtained using local learning rules that adjust the weights in the network based on the past interactions between the nodes. We conducted a detailed simulation study of this learning rule, which is reminiscent of temporal delay reinforcement learning. We show that such type of learning increases the accessibility of all paths on the network and allows such complex systems to overcome engrained structure/paths and allows it to adapt to new challenges. This work has been submitted to IEEE Transactions on Network Science and Engineering. We were also involved in two projects that relate to Statistical and Machine Learning, but not reported under the Statistical Learning for Biomedical project (with James Malley as LI). The first involved using machine learning and the large genetic and other data available publicly for the C. elegans worm to investigate to what degree the known neural connectivity of C. elegans can be estimated from the genetic, geometric and other information that is available on various public databases. We showed that even with a limited set of gene expression data that we gathered, accuracy of 86% percent can be achieved. We expect the accuracy to improve as more genetic data become available in the future. This work was presented at the Annual Meeting of the Society for Neuroscience, San Diego, in November of 2013. The second project involved using machine learning to estimate behavioral modes of mouse to be used with The System for Continuous Observation of Rodents in Home-cage Environment (SCORHE). This work was published in Behavior Research Methods in 2014. A project with investigators from the Laboratory of Clinical and Developmental Genomics, NICHD we study neuronal cultures created by reprogramming skin cells from autistic patients as well as normals. It is a part of a larger study addressing the molecular and cellular changes that occur in the autistic brain during the development. In the current study the goal is to identify the changes in the network structure, or in the activity profile of different cells that distinguishes the normal cell cultures from those of autistic patients. In a continuing project with investigators in the Program on Pediatric Imaging and Tissue Sciences (PPITS), NICHD we conduct a theoretical study of the observed skewed and heavy-tailed axon diameter distribution and arrive at a parametric form that optimizes the informative upper bound (IUB) as well as the information capacity of the nerve fascicles. A manuscript describing this work is published in PLOS ONE in 2013. In a follow-up work, this distribution has been implemented and applied to simulated and experimental data, yielding improved measurements of the axon diameter distributions.