Search

The experimental and computational methods were used to assess the applicability of the mixed-genome microarray (MGM) in determining phylogenetic relationships among strains of bacteria.

Two conformal finite-difference time-domain (FDTD) methods are considered, the contour path (CPFDTD) method of Jurgens et al. [4] and the overlapping grid (OGFDTD) method of Yee et al. [6]. Both TE and TM scattering from a two-dimensional (2-D), perfectly conducting circular cylinder are used to test the accuracy of the methods for curved surfaces. Also, TE and TM scattering from a 2-D, perfectly-conducting rotated square cylinder are used to test the accuracy for corners and edges. It is shown that the conformal method proposed by Yee et al. provide significant improvement in accuracy over the original FDTD algorithm for most of the geometries studied in this paper. However, implementation becomes more difficult as geometries become more complex. The conformal method proposed by Jurgens et al. provide significant improvement in accuracy as well for most of the geometries studied in this paper. However, improvement does not occur for the TM case when the square cylinder is not aligned properly with the grid. Implementation of the CPFDTD method is relatively straightforward. For the majority of the cases studied, the OGFDTD method is more accurate than the CPFDTD method.

A Monte-Carlo finite-difference time-domain (FDTD) technique is developed for wave scattering from randomly rough, one-dimensional surfaces satisfying the Dirichlet boundary condition. Both single-scale Gaussian and multiscale Pierson-Moskowitz surface roughness spectra are considered. Bistatic radar cross sections are calculated as a function of scattering angle for incident angles of 0, 45, 70, and 80 degrees measured from the vertical. The contour path FDTD method is shown to improve accuracy for incident angles greater than 45 degrees. Results compare well with those obtained using a Monte-Carlo integral equation technique.

Bistatic radar cross sections are calculated using two modern scattering models: the small slope approximation (both first- and second-order) and the phase perturbation technique. The problem is limited to scalar-wave scattering from two-dimensional, randomly rough Dirichlet surfaces with a Gaussian roughness spectrum. Numerical results for the cross sections are compared to those found using the classical Kirchhoff, or physical optics, approximation and perturbation theory. Over a wide range of scattering angles, the new results agree well with the classical results when the latter are considered to be accurate. A comparison between the new results shows that the phase perturbation method gives better results in the backscattering region for correlation lengths greater than approximately one wavelength, while both the first- and second-order small slope approximations yield greater accuracy in the forward scattering direction at low grazing angles.

Abstract Background Classification microarrays are used for purposes such as identifying strains of bacteria and determining genetic relationships to understand the epidemiology of an infectious disease. For these cases, mixed microarrays, which are composed of DNA from more than one organism, are more effective than conventional microarrays composed of DNA from a single organism. Selection of probes is a key factor in designing successful mixed microarrays because redundant sequences are inefficient and limited representation of diversity can restrict application of the microarray. We have developed a Java-based software tool, called PLASMID, for use in selecting the minimum set of probe sequences needed to classify different groups of plasmids or bacteria. Results The software program was successfully applied to several different sets of data. The utility of PLASMID was illustrated using existing mixed-plasmid microarray data as well as data from a virtual mixed-genome microarray constructed from different strains of Streptococcus. Moreover, use of data from expression microarray experiments demonstrated the generality of PLASMID. Conclusion In this paper we describe a new software tool for selecting a set of probes for a classification microarray. While the tool was developed for the design of mixed microarrays–and mixed-plasmid microarrays in particular–it can also be used to design expression arrays. The user can choose from several clustering methods (including hierarchical, non-hierarchical, and a model-based genetic algorithm), several probe ranking methods, and several different display methods. A novel approach is used for probe redundancy reduction, and probe selection is accomplished via stepwise discriminant analysis. Data can be entered in different formats (including Excel and comma-delimited text), and dendrogram, heat map, and scatter plot images can be saved in several different formats (including jpeg and tiff). Weights generated using stepwise discriminant analysis can be stored for analysis of subsequent experimental data. Additionally, PLASMID can be used to construct virtual microarrays with genomes from public databases, which can then be used to identify an optimal set of probes.

Advances in genome sequencing technology and computing power have brought about the explosive growth of sequenced genomes in public repositories with a concomitant increase in annotation errors. Many protein sequences are annotated using computational analysis rather than experimental verification, leading to inaccuracies in annotation. Confirmation of existing protein annotations is urgently needed before misannotation becomes even more prevalent due to error propagation. In this work we present a novel approach for automatically confirming the existence of manually curated information with experimental evidence of protein annotation. Our ensemble learning method uses a combination of recurrent convolutional neural network, logistic regression, and support vector machine models. Natural language processing in the form of word embeddings is used with journal publication titles retrieved from the UniProtKB database. Importantly, we use recall as our most significant metric to ensure the maximum number of verifications possible; results are reported to a human curator for confirmation. Our ensemble model achieves 91.25% recall, 71.26% accuracy, 65.19% precision, and an F1 score of 76.05% and outperforms the Bidirectional Encoder Representations from Transformers for Biomedical Text Mining (BioBERT) model with fine-tuning using the same data. [ABSTRACT FROM AUTHOR]

The lowest-order nonlocal small slope approximation (NLSSA) reflection coefficient is derived, and numerical results are presented for one-dimensional (1-D) surfaces, satisfying the Dirichlet boundary condition. Reduction to the perturbation expression occurs in the small height limit, and the first term of the reflection coefficient gives the Kirchhoff approximation (KA) result. Numerical results for the reflection loss at low grazing angles using a Pierson-Moskowitz spectrum are compared with those of other approximate methods as well as with exact integral equation (IE) results. For those cases when exact results are available, the NLSSA is found to give accurate results, comparable to those of the small slope approximation (SSA) and superior to those of classical perturbation theory (PT) and the KA. Index Terms - Acoustic scattering, electromagnetic reflection, electromagnetic scattering by rough surfaces, radar scattering, remote sensing, sea surface electromagnetic scattering.