The properties of certain networks are determined by hidden variables that are not explicitly measured. The conditional probability (propagator) that a vertex with a given value of the hidden variable is connected to k of other vertices determines all measurable properties. We study hidden variable models and find an averaging approximation that enables us to obtain a general analytical result for the propagator. Analytic results showing the validity of the approximation are obtained. We apply hidden variable models to protein-protein interaction networks (PINs) in which the hidden variable is the association free-energy, determined by distributions that depend on biochemistry and evolution. We compute degree distributions as well as clustering coefficients of several PINs of different species; good agreement with measured data is obtained. For the human interactome two different parameter sets give the same degree distributions, but the computed clustering coefficients differ by a factor of about two. This shows that degree distributions are not sufficient to determine the properties of PINs.
Conformation Clustering of Long MD Protein Dynamics with an Adversarial Autoencoder
The Johns Hopkins University(约翰斯·霍普金斯大学)
近年来,计算机硬件的发展极大地促进了原子级分子动力学模拟的发展。单个GPU连接的集群能够在合理的时间内产生微秒长度的轨迹。多个蛋白质的状态和大量的微观状态与折叠和功能的蛋白质可以观察到的构象采样的轨迹。然而,聚类这些构象是识别蛋白质状态,评估转换速率和理解蛋白质行为所必需的。在本文中,我们提出了一种新的数据驱动的生成式构象聚类方法的基础上的对抗性自动编码器(AAE),并提供了相关的软件实现Cong。使用从D.E. Shaw Research Group.所提出的聚类算法确定了许多折叠过程中的显着特点,通过分组的构象,共享不容易识别的轨迹中的共同特征。
One-step, Wash-free, Nanoparticle Clustering-based Magnetic Particle Spectroscopy (MPS) Bioassay Method for Detection of SARS-CoV-2 Spike and Nucleocapsid Proteins in Liquid Phase
Subcellular proteome niche discovery using semi-supervised functional clustering
University of California, Santa Cruz(加利福尼亚大学圣克鲁兹分校);Marine Chemistry and Geochemistry Department, Woods Hole Oceanographic Institution(伍兹霍尔海洋研究所);Department of Biology, Mount Allison University(蒙茅斯阿尔金大学)
Bayesian analysis of data from segmented super-resolution images for quantifying protein clustering
超分辨率成像技术大大提高了我们在生物系统中可视化纳米结构的能力。它们的应用进一步使人们能够潜在地定量相关参数以确定细胞中的分子组织和化学计量。然而,荧光发射和标记策略的固有随机性迫使使用专用方法来精确测量这些参数。在这里,我们描述了一种贝叶斯方法来精确定量的相对丰度的分子寡聚体从分割的图像。集群中分子数量的代理分布-例如定位数量或荧光强度-通过嵌套采样算法进行拟合,以比较复杂性不断增加的混合模型,并确定混合组分的最佳数量及其权重。我们测试了该算法的性能{ \it in silico}数据作为数据点的数量,阈值和分布形状的函数。我们将这些结果与其他统计方法获得的结果进行比较,显示了我们方法的改进性能。我们的方法提供了一个强大的工具,在拟合从荧光成像中提取的数据的模型选择,从而提高了参数确定的精度。重要的是,这种方法的最大好处发生在小统计量或不完整的数据集上,从而能够在单个图像水平上进行准确的分析。我们进一步提出了其应用的结果,从超分辨率成像的HeLa细胞中的动力蛋白的实验数据,证实了细胞质的单电机和高阶结构的混合人口的存在。