Wrapper-based selection of genetic features in genome-wide association studies through fast matrix operations

ABSTRACT: BACKGROUND: Through the wealth of information contained within them, genome-wide association studies (GWAS) have the potential to provide researchers with a systematic means of associating genetic variants with a wide variety of disease phenotypes. Due to the limitations of approaches that have analyzed single variants one at a time, it has been proposed that the genetic basis of these disorders could be determined through detailed analysis of the genetic variants themselves and in conjunction with one another. The construction of models that account for these subsets of variants requires methodologies that generate predictions based on the total risk of a particular group of polymorphisms. However, due to the excessive number of variants, constructing these types of models has so far been computationally infeasible. RESULTS: We have implemented an algorithm, known as greedy RLS, that we use to perform the first known wrapper-based feature selection on the genome-wide level. The running time of greedy RLS grows linearly in the number of training examples, the number of features in the original data set, and the number of selected features. This speed is achieved through computational short-cuts based on matrix calculus. Since the memory consumption in present-day computers can form an even tighter bottleneck than running time, we also developed a space efficient variation of greedy RLS which trades running time for memory. These approaches are then compared to traditional wrapper-based feature selection implementations based on support vector machines (SVM) to reveal the relative speed-up and to assess the feasibility of the new algorithm. As a proof of concept, we apply greedy RLS to the Hypertension - UK National Blood Service WTCCC dataset and select the most predictive variants using 3-fold external cross-validation in less than 26 minutes on a high end desktop. On this dataset, we also show that greedy RLS has a better classification performance on independent test data than a classifier trained using features selected by a statistical p-value-based filter, which is currently the most popular approach for constructing predictive models in GWAS. CONCLUSIONS: Greedy RLS is the first known implementation of a machine learning based method with the capability to conduct a wrapper-based feature selection on an entire GWAS containing several thousand examples and over 400,000 variants. In our experiments, greedy RLS selected a highly predictive subset of genetic variants in a fraction of the time spent by wrapper-based selection methods used together with SVM classifiers. The proposed algorithms are freely available as part of the RLScore software library at http://users.utu.fi/aatapa/RLScore/.     

Exchange of CO2, CH4 and N2O between the atmosphere and two northern boreal ponds with catchments dominated by peatlands or forests

Concentrations of carbon dioxide (CO₂), methane (CH₄) and nitrous oxide (N₂O) in the water column and their exchange at the water/air interface were studied during the open water period in two freshwater ponds with different catchment characteristics in the northern boreal zone in Finland; either peatlands or coniferous upland forests dominated the catchment of the ponds. Both ponds were supersaturated with dissolved CO₂ and CH₄ with respect to the equilibrium with the atmosphere, but were close to the equilibrium with N₂O. The mean CO₂ efflux from the pond was higher in the peatland-dominated catchment (22 mg m^–2 h^–1) than in the forested catchment (0.7 mg m^–2 h^–1), whereas the mean CH₄ emissions were similar (7.6 and 3.5 mg m^–2 d^–1, respectively). The fluxes of N₂O were generally negligible. The higher CO₂ concentrations and efflux in the pond with the peatland-dominated catchment were attributed to a greater input of allochthonous carbon to that pond from its catchment due to its higher water colour and higher total organic carbon (TOC) concentration. The water pH, which also differed between the ponds, could additionally affect the CO₂ dynamics. Since the catchment characteristics can regulate aquatic carbon cycles, catchment-scale studies are needed to attain a deeper understanding of the aquatic greenhouse gas dynamics.     

Particle Size, Moisture, and Fluidization Variations Described by Indirect In-line Physical Measurements of Fluid Bed Granulation

The aim of this study was to evaluate an instrumentation system for a bench scale fluid bed granulator to determine the parameters expressing the changing conditions during the spraying phase of a fluid bed process. The study focused mainly on four in-line measurements (dependent variables): fluidization parameter (calculated by inlet air flow rate and rotor speed), pressure difference over the upper filters, pressure difference over the granules (lower filter), and temperature of the fluidizing mass. In-line particle size measured by the spatial filtering technique was an essential predictor variable. Other physical process measurements of the automated granulation system, 25 direct and 12 derived parameters, were also utilized for multivariate modeling. The correlation and partial least squares analyses revealed significant relationships between various process parameters highlighting the particle size, moisture, and fluidization effect. Fluidization parameter and pressure difference over upper filters were found to correlate with in-line particle size and therefore could be used as estimates of particle size during granulation. The pressure difference over the granules and the temperature of the fluidizing mass expressed the moisture conditions of wet granulation. The instrumentation system evaluated here is an invaluable aid to gaining more control for fluid bed processing to obtain repeatable granules for further processing.     

A New Rapid On-Line Imaging Method to Determine Particle Size Distribution of Granules

The purpose of this research was to study the feasibility of the new image analysis method in the particle size determination of the granules. The method is capable of forming a three-dimensional topographic image of a sample surface from a digital picture. In the method, a flat granule bed surface was illuminated from three different directions, using the three primary colors (red, green, and blue). One color picture was taken by a digital camera, after which a topographic image of the object surface was constructed. The particle size distribution was then calculated from the image data. The particle size analysis method was tested both off-line and on-line. Off-line particle size measurement results determined by the image analysis method corresponded quite well to those of sieve analysis in the size fraction range 250–1,000 μm. In on-line application, images were successfully retrieved and median granule size trend could be calculated and followed during fluid bed granulations.     

Effects of metabolic rate on thermal responses at different air velocities in −10°C

The effects of exercise intensity on thermoregulatory responses in cold (−10°C) in a 0.2 (still air, NoWi), 1.0 (Wi1), and 5.0 (Wi5) m s⁻¹ wind were studied. Eight young and healthy men, preconditioned in thermoneutral (+20°C) environment for 60 min, walked for 60 min on the treadmill at 2.8 km/h with different combinations of wind and exercise intensity. Exercise level was adjusted by changing the inclination of the treadmill between 0° (lower exercise intensity, metabolic rate 124 W m⁻², LE) and 6° (higher exercise intensity, metabolic rate 195 W m⁻², HE). Due to exercise increased heat production and circulatory adjustments, the rectal temperature (T_re), mean skin temperature (T_sk) and mean body temperature (T_b) were significantly higher at the end of HE in comparison to LE in NoWi and Wi1, and T_re and T_b also in Wi5. T_sk and T_b were significantly decreased by 5.0 m s⁻¹ wind in comparison to NoWi and Wi1. The higher exercise intensity was intense enough to diminish peripheral vasoconstriction and consequently the finger skin temperature was significantly higher at the end of HE in comparison to LE in NoWi and Wi1. Mean heat flux from the skin was unaffected by the exercise intensity. At LE oxygen consumption (V ₂) was significantly higher in Wi5 than NoWi and Wi1. Heart rate was unaffected by the wind speed. The results suggest that, with studied exercise intensities, produced without changes in walking speed, the metabolic rate is not so important that it should be taken into consideration in the calculation of wind chill index.     

Fimbrial Subunit Protein FaeG Expressed in Transgenic Tobacco Inhibits the Binding of F4ac Enterotoxigenic Escherichia coli to Porcine Enterocytes

Plants offer a promising alternative for the production of foreign proteins for pharmaceutical purposes in tissues that are consumed as food and/or feed. Our long-term strategy is to develop edible vaccines against piglet diarrhoea caused by enterotoxigenic Escherichia coli (F4 ETEC) in feed plants. In this work, we isolated a gene, faeG, encoding for a major F4ac fimbrial subunit protein. Our goal was to test whether the FaeG protein, when isolated from its fimbrial background and produced in a plant cell, would retain the key properties of an oral vaccine, that is, stability in gastrointestinal conditions, binding to intestinal receptors and inhibition of the F4 ETEC attachment. For this purpose, tobacco was first transformed with a faeG construct that included a transit peptide encoding sequence to target the FaeG protein to the chloroplast. The best transgenic lines produced FaeG protein in amounts of 1% total soluble protein. The stability of the plant-produced FaeG was tested in fluids simulating piglet gastric (SGF) and intestinal (SIF) conditions. Plant-produced FaeG proved to be stable up to 2 h under these conditions. The binding and inhibition properties were tested with isolated piglet villi. These results showed that the plant-produced FaeG could bind to the receptors on the villi and subsequently inhibit F4 ETEC binding in a dose-dependent manner. Thus, the first two prerequisites for the development of an oral vaccine have been met.     

Regulation of the sequential processing of Semliki Forest virus replicase polyprotein

The replication of most positive-strand RNA viruses and retroviruses is regulated by proteolytic processing. Alphavirus replicase proteins are synthesized as a polyprotein, called P1234, which is cleaved into nsP1, nsP2, nsP3, and nsP4 by the carboxyl-terminal protease domain of nsP2. The cleavage intermediate P123+nsP4 synthesizes minus-strand copies of the viral RNA genome, whereas the completely processed complex is required for plus-strand synthesis. To understand the mechanisms responsible for this sequential proteolysis, we analyzed in vitro translated Semliki Forest virus polyproteins containing noncleavable processing sites or various deletions. Processing of each of the three sites in vitro required a different type of activity. Site 3/4 was cleaved in trans by nsP2, its carboxyl-terminal fragment Pro39, and by all polyprotein proteases. Site 1/2 was cleaved in cis with a half-life of about 20-30 min. Site 2/3 was cleaved rapidly in trans but only after release of nsP1 from the polyprotein exposing an "activator" sequence present in the amino terminus of nsP2. Deletion of amino-terminal amino acids of nsP2 or addition of extra amino acid residues to its amino terminus specifically inhibited the protease activity that processes the 2/3 site. This sequence of delayed processing of P1234 would explain the accumulation of P123 plus nsP4, the early short-lived minus-strand replicase. The polyprotein stage would allow correct assembly and membrane association of the RNA-polymerase complex. Late in infection free nsP2 would cleave at site 2/3 yielding P12 and P34, the products of which, nsP1-4, are distributed to the plasma membrane, nucleus, cytoplasmic aggregates, and proteasomes, respectively.     

A new microvolume technique for bioaffinity assays using two-photon excitation

Bioaffinity binding assays such as the immunoassay are widely used in life science research. In an immunoassay, specific antibodies are used to bind target molecules in the sample, and quantification of the binding reaction reveals the amount of the target molecules. Here we present a method to measure bioaffinity assays using the two-photon excitation of fluorescence. In this method, microparticles are used as solid phase in binding the target molecules. The degree of binding is then quantified from individual microparticles by use of two photon excitation of fluorescence. We demonstrated the effectiveness of the method using the human alpha-fetoprotein (AFP) immunoassay, which is used to detect fetal disorders. The sensitivity and dynamic range we obtained with this assay indicate that this method can provide a cost-effective and simple way to measure various biomolecules in solution for research and clinical applications.     

Tight junction proteins in gallbladder epithelium: different expression in acute acalculous and calculous cholecystitis.

There is a paucity of information of tight junction (TJ) proteins in gallbladder epithelium, and disturbances in the structure of these proteins may play a role in the pathogenesis of acute acalculous cholecystitis (AAC) and acute calculous cholecystitis (ACC). Using immunohistochemistry, we investigated the expression of TJ proteins claudin-1, -2, -3, and -4, occludin, zonula occludens (ZO-1), and E-cadherin in 9 normal gallbladders, 30 gallbladders with AAC, and 21 gallbladders with ACC. The number of positive epithelial and endothelial cells and the intensity of the immunoreaction were determined. Membrane-bound and cytoplasmic immunoreactivities were separately assessed. We found that TJ proteins were uniformly expressed in normal gallbladder epithelium, with the exception of claudin-2, which was present in less than half of the cells. In AAC, expression of cytoplasmic occludin and claudin-1 were decreased, as compared with normal gallbladder. In ACC, expression of claudin-2 was increased, and expression of claudin-1, -3, and -4, occludin, and ZO-1 were decreased, as compared with normal gallbladder or AAC. We conclude that there are significant differences in expression of TJ proteins in AAC and ACC, supporting the idea that AAC represents a manifestation of systemic inflammatory disease, whereas ACC is a local inflammatory and often infectious disease.