Inference of genetic regulatory networks with regulatory hubs using vector autoregressions and automatic relevance determination with model selections

The inference of genetic regulatory networks (GRNs) reveals how genes interact with each other. A few genes can regulate many genes as targets to control cell functions. We present new methods based on the order-1 vector autoregression (VAR1) for inferring GRNs from gene expression time series. The methods use the automatic relevance determination (ARD) to incorporate the regulatory hub structure into the estimation of VAR1 in a Bayesian framework. Several sparse approximation schemes are applied to the estimated regression weights or VAR1 model to generate the sparse weighted adjacency matrices representing the inferred GRNs. We apply the proposed and several widespread reference methods to infer GRNs with up to 100 genes using simulated, DREAM4 in silico and experimental E. coli gene expression time series. We show that the proposed methods are efficient on simulated hub GRNs and scale-free GRNs using short time series simulated by VAR1s and outperform reference methods on small-scale DREAM4 in silico GRNs and E. coli GRNs. They can utilize the known major regulatory hubs to improve the performance on larger DREAM4 in silico GRNs and E. coli GRNs. The impact of nonlinear time series data on the performance of proposed methods is discussed.

Scalable Bayesian GPFA with automatic relevance determination and discrete noise models

Latent variable models are ubiquitous in the exploratory analysis of neural population recordings, where they allow researchers to summarize the activity of large populations of neurons in lower dimensional 'latent' spaces. Existing methods can generally be categorized into (i) Bayesian methods that facilitate flexible incorporation of prior knowledge and uncertainty estimation, but which typically do not scale to large datasets; and (ii) highly parameterized methods without explicit priors that scale better but often struggle in the low-data regime. Here, we bridge this gap by developing a fully Bayesian yet scalable version of Gaussian process factor analysis (bGPFA) which models neural data as arising from a set of inferred latent processes with a prior that encourages smoothness over time. Additionally, bGPFA uses automatic relevance determination to infer the dimensionality of neural activity directly from the training data during optimization. To enable the analysis of continuous recordings without trial structure, we introduce a novel variational inference strategy that scales near-linearly in time and also allows for non-Gaussian noise models more appropriate for electrophysiological recordings. We apply bGPFA to continuous recordings spanning 30 minutes with over 14 million data points from primate motor and somatosensory cortices during a self-paced reaching task. We show that neural activity progresses from an initial state at target onset to a reach-specific preparatory state well before movement onset. The distance between these initial and preparatory latent states is predictive of reaction times across reaches, suggesting that such preparatory dynamics have behavioral relevance despite the lack of externally imposed delay periods. Additionally, bGPFA discovers latent processes that evolve over slow timescales on the order of several seconds and contain complementary information about reaction time. These timescales are longer than those revealed by methods which focus on individual movement epochs and may reflect fluctuations in e.g. task engagement.

Parameterization of wind evolution using lidar

Wind evolution, i.e., the evolution of turbulence structures over time, has become an increasingly interesting topic in recent years, mainly due to the development of lidar-assisted wind turbine control, which requires accurate prediction of wind evolution to avoid unnecessary or even harmful control actions. Moreover, 4D stochastic wind field simulations can be made possible by integrating wind evolution into standard 3D simulations to provide a more realistic simulation environment for this control concept. Motivated by these factors, this research aims to investigate the potential of Gaussian process regression in the parameterization of wind evolution. Wind evolution is commonly quantified using magnitude-squared coherence of wind speed and is estimated with lidar data measured by two nacelle-mounted lidars in this research. A two-parameter wind evolution model modified from a previous study is used to model the estimated coherence. A statistical analysis is done for the wind evolution model parameters determined from the estimated coherence to provide some insights into the characteristics of wind evolution. Gaussian process regression models are trained with the wind evolution model parameters and different combinations of wind-field-related variables acquired from the lidars and a meteorological mast. The automatic relevance determination squared exponential kernel function is applied to select suitable variables for the models. The performance of the Gaussian process regression models is analyzed with respect to different variable combinations, and the selected variables are discussed to shed light on the correlation between wind evolution and these variables.

Prediction of the space group of perovskite-like compounds of the composition AII2BIIIB′VO6

The prediction of new compounds of the composition AII2BIIIB′VO6 was carried out, the type of distortion of their perovskite-like lattice, the space group were predicted, and the parameters of the crystal lattice of the predicted compounds were estimated. When predicting, only the property values of the chemical elements were used. Programs based on machine learning algorithms for various variants of neural networks, a linear machine, the formation of logical regularities, k-nearest neighbors, support vector machine showed the best results when predicting the type of distortion of a perovskite-like lattice. When evaluating the lattice parameters, the programs based on algorithms for orthogonal matching pursuit and automatic relevance determination regression were the most accurate methods. The accuracy of predictions of the perovskite-like lattice distortion type was no less than 74 %. The accuracy of estimating the lattice linear parameters was within ± 0.0120 – 0.8264 Å, and the accuracy for angles β with monoclinic distortion of the lattice was ± 0.08 – 0.74 deg. The calculations were carried out using systems based on machine learning methods. To evaluate the prediction accuracy, an exam recognition in the cross-validation mode was used for the compounds included in the sample for machine learning. The predicted compounds are promising for the search for new magnetic, thermoelectric and dielectric materials.