Uncertainty Quantification with Support Vector Regression Prediction Models

Uncertainty quantification of petroleum reservoir models is one of the present challenges, which is usually approached with a wide range of geostatistical tools linked with statistical optimisation or/and inference algorithms. The paper considers a data driven approach in modelling uncertainty in spatial predictions. Proposed semi-supervised Support Vector Regression (SVR) model has demonstrated its capability to represent realistic features and describe stochastic variability and non-uniqueness of spatial properties. It is able to capture and preserve key spatial dependencies such as connectivity, which is often difficult to achieve with two-point geostatistical models. Semi-supervised SVR is designed to integrate various kinds of conditioning data and learn dependences from them. A stochastic semi-supervised SVR model is integrated into a Bayesian framework to quantify uncertainty with multiple models fitted to dynamic observations. The developed approach is illustrated with a reservoir case study. The resulting probabilistic production forecasts are described by uncertainty envelopes.

Use of human MAR elements to improve retroviral vector production.

Retroviral vectors have many favorable properties for gene therapies, but their use remains limited by safety concerns and/or by relatively lower titers for some of the safer self-inactivating (SIN) derivatives. In this study, we evaluated whether increased production of SIN retroviral vectors can be achieved from the use of matrix attachment region (MAR) epigenetic regulators. Two MAR elements of human origin were found to increase and to stabilize the expression of the green fluorescent protein transgene in stably transfected HEK-293 packaging cells. Introduction of one of these MAR elements in retroviral vector-producing plasmids yielded higher expression of the viral vector RNA. Consistently, viral titers obtained from transient transfection of MAR-containing plasmids were increased up to sixfold as compared with the parental construct, when evaluated in different packaging cell systems and transfection conditions. Thus, use of MAR elements opens new perspectives for the efficient generation of gene therapy vectors.

An adenovirus-based fluorescent reporter vector to identify and isolate HIV-infected cells.

A procedure is described that allows the simple identification and sorting of live human cells that transcribe actively the HIV virus, based on the detection of GFP fluorescence in cells. Using adenoviral vectors for gene transfer, an expression cassette including the HIV-1 LTR driving the reporter gene GFP was introduced into cells that expressed stably either the Tat transcriptional activator, or an inactive mutant of Tat. Both northern and fluorescence-activated cell sorting (FACS) analysis indicate that cells containing the functional Tat protein presented levels of GFP mRNA and GFP fluorescence several orders of magnitude higher than control cells. Correspondingly, cells infected with HIV-1 showed similar enhanced reporter gene activation. HIV-1-infected cells of the lymphocytic line Jurkat were easily identified by fluorescence-activated cell sorting (FACS) as they displayed a much higher green fluorescence after transduction with the reporter adenoviral vector. This procedure could also be applied on primary human cells as blood monocyte-derived macrophages exposed to the adenoviral LTR-GFP reporter presented a much higher fluorescence when infected with HIV-1 compared with HIV-uninfected cells. The vector described has the advantages of labelling cells independently of their proliferation status and that analysis can be carried on intact cells which can be isolated subsequently by fluorescence-activated cell sorting (FACS) for further culture. This work suggests that adenoviral vectors carrying a virus-specific transcriptional control element controlling the expressions of a fluorescent protein will be useful in the identification and isolation of cells transcribing actively the viral template, and to be of use for drug screening and susceptibility assays.

Geomodelling of a fluvial system with semi-supervised support vector regression

Fluvial deposits are a challenge for modelling flow in sub-surface reservoirs. Connectivity and continuity of permeable bodies have a major impact on fluid flow in porous media. Contemporary object-based and multipoint statistics methods face a problem of robust representation of connected structures. An alternative approach to model petrophysical properties is based on machine learning algorithm ? Support Vector Regression (SVR). Semi-supervised SVR is able to establish spatial connectivity taking into account the prior knowledge on natural similarities. SVR as a learning algorithm is robust to noise and captures dependencies from all available data. Semi-supervised SVR applied to a synthetic fluvial reservoir demonstrated robust results, which are well matched to the flow performance

Wave-vector dependence of spin and density multipole excitations in quantum dots

We have employed time-dependent local-spin density-functional theory to analyze the multipole spin and charge density excitations in GaAs-AlxGa1-xAs quantum dots. The on-plane transferred momentum degree of freedom has been taken into account, and the wave-vector dependence of the excitations is discussed. In agreement with previous experiments, we have found that the energies of these modes do not depend on the transferred wave vector, although their intensities do. Comparison with a recent resonant Raman scattering experiment [C. Schüller et al., Phys. Rev. Lett. 80, 2673 (1998)] is made. This allows us to identify the angular momentum of several of the observed modes as well as to reproduce their energies

Landscape structure affects dispersal in the greater white-toothed shrew: inference between genetic and simulated ecological distances

Animal dispersal in a fragmented landscape depends on the complex interaction between landscape structure and animal behavior. To better understand how individuals disperse, it is important to explicitly represent the properties of organisms and the landscape in which they move. A common approach to modelling dispersal includes representing the landscape as a grid of equal sized cells and then simulating individual movement as a correlated random walk. This approach uses a priori scale of resolution, which limits the representation of all landscape features and how different dispersal abilities are modelled. We develop a vector-based landscape model coupled with an object-oriented model for animal dispersal. In this spatially explicit dispersal model, landscape features are defined based on their geographic and thematic properties and dispersal is modelled through consideration of an organism's behavior, movement rules and searching strategies (such as visual cues). We present the model's underlying concepts, its ability to adequately represent landscape features and provide simulation of dispersal according to different dispersal abilities. We demonstrate the potential of the model by simulating two virtual species in a real Swiss landscape. This illustrates the model's ability to simulate complex dispersal processes and provides information about dispersal such as colonization probability and spatial distribution of the organism's path

Generation of a tightly regulated all-cis beta cell-specific tetracycline-inducible vector.

Ability to induce protein expression at will in a cell is a powerful strategy used by scientists to better understand the function of a protein of interest. Various inducible systems have been designed in eukaryotic cells to achieve this goal. Most of them rely on two distinct vectors, one encoding a protein that can regulate transcription by binding a compound X, and one hosting the cDNA encoding the protein of interest placed downstream of promoter sequences that can bind the protein regulated by compound X (e.g., tetracycline, ecdysone). The commercially available systems are not designed to allow cell- or tissue-specific regulated expression. Additionally, although these systems can be used to generate stable clones that can be induced to express a given protein, extensive screening is often required to eliminate the clones that display poor induction or high basal levels. In the present report, we aimed to design a pancreatic beta cell-specific tetracycline-inducible system. Since the classical two-vector based tetracycline-inducible system proved to be unsatisfactory in our hands, a single vector was eventually designed that allowed tight beta cell-specific tetracycline induction in unselected cell populations.

Surface granularity as a discriminating feature of illicit tablets

In this paper we propose an innovative methodology for automated profiling of illicit tablets bytheir surface granularity; a feature previously unexamined for this purpose. We make use of the tinyinconsistencies at the tablet surface, referred to as speckles, to generate a quantitative granularity profileof tablets. Euclidian distance is used as a measurement of (dis)similarity between granularity profiles.The frequency of observed distances is then modelled by kernel density estimation in order to generalizethe observations and to calculate likelihood ratios (LRs). The resulting LRs are used to evaluate thepotential of granularity profiles to differentiate between same-batch and different-batches tablets.Furthermore, we use the LRs as a similarity metric to refine database queries. We are able to derivereliable LRs within a scope that represent the true evidential value of the granularity feature. Thesemetrics are used to refine candidate hit-lists form a database containing physical features of illicittablets. We observe improved or identical ranking of candidate tablets in 87.5% of cases when granularityis considered.

Learning wind fields with multiple kernels

This paper presents multiple kernel learning (MKL) regression as an exploratory spatial data analysis and modelling tool. The MKL approach is introduced as an extension of support vector regression, where MKL uses dedicated kernels to divide a given task into sub-problems and to treat them separately in an effective way. It provides better interpretability to non-linear robust kernel regression at the cost of a more complex numerical optimization. In particular, we investigate the use of MKL as a tool that allows us to avoid using ad-hoc topographic indices as covariables in statistical models in complex terrains. Instead, MKL learns these relationships from the data in a non-parametric fashion. A study on data simulated from real terrain features confirms the ability of MKL to enhance the interpretability of data-driven models and to aid feature selection without degrading predictive performances. Here we examine the stability of the MKL algorithm with respect to the number of training data samples and to the presence of noise. The results of a real case study are also presented, where MKL is able to exploit a large set of terrain features computed at multiple spatial scales, when predicting mean wind speed in an Alpine region.