Michael L. Raymer

Dimensionality reduction using genetic algorithms

Pattern recognition generally requires that objects be described in terms of a set of measurable features. The selection and quality of the features representing each pattern affect the success of subsequent classification. Feature extraction is the process of deriving new features from original features to reduce the cost of feature measurement, increase classifier efficiency, and allow higher accuracy. Many feature extraction techniques involve linear transformations of the original pattern vectors to new vectors of lower dimensionality. While this is useful for data visualization and classification efficiency, it does not necessarily reduce the number of features to be measured since each new feature may be a linear combination of all of the features in the original pattern vector. Here, we present a new approach to feature extraction in which feature selection and extraction and classifier training are performed simultaneously using a genetic algorithm. The genetic algorithm optimizes a feature weight vector used to scale the individual features in the original pattern vectors. A masking vector is also employed for simultaneous selection of a feature subset. We employ this technique in combination with the k nearest neighbor classification rule, and compare the results with classical feature selection and extraction techniques, including sequential floating forward feature selection, and linear discriminant analysis. We also present results for the identification of favorable water-binding sites on protein surfaces.

Empirical Analysis of the STR Profiles Resulting from Conceptual Mixtures

Samples containing DNA from two or more individuals can be difficult to interpret. Even ascertaining the number of contributors can be challenging and associated uncertainties can have dramatic effects on the interpretation of testing results. Using an FBI genotypes dataset, containing complete genotype information from the 13 Combined DNA Index System (CODIS) loci for 959 individuals, all possible mixtures of three individuals were exhaustively and empirically computed. Allele sharing between pairs of individuals in the original dataset, a randomized dataset and datasets of generated cousins and siblings was evaluated as were the number of loci that were necessary to reliably deduce the number of contributors present in simulated mixtures of four or less contributors. The relatively small number of alleles detectable at most CODIS loci and the fact that some alleles are likely to be shared between individuals within a population can make the maximum number of different alleles observed at any tested loci an unreliable indicator of the maximum number of contributors to a mixed DNA sample. This analysis does not use other data available from the electropherograms (such as peak height or peak area) to estimate the number of contributors to each mixture. As a result, the study represents a worst case analysis of mixture characterization. Within this dataset, approximately 3% of three-person mixtures would be mischaracterized as two-person mixtures and more than 70% of four-person mixtures would be mischaracterized as two- or three-person mixtures using only the maximum number of alleles observed at any tested locus.

Work in progress - the WSU model for engineering mathematics education

This paper summarizes progress to date on the WSU model for engineering mathematics education, an NSF funded curriculum reform initiative at Wright State University. The WSU model seeks to increase student retention, motivation and success in engineering through application-driven, just-in-time engineering math instruction. The WSU approach involves the development of a novel freshman-level engineering mathematics course EGR 101, as well as a large-scale restructuring of the engineering curriculum. By removing traditional math prerequisites and moving core engineering courses earlier in the program, the WSU model shifts the traditional emphasis on math prerequisite requirements to an emphasis on engineering motivation for math, with a just-in-time structuring of the new math sequence. This paper summarizes the development to date of the WSU model for engineering mathematics education, including a preliminary assessment of student performance and perception during the initial implementation of EGR 101. In addition, an assessment of first-year retention results is anticipated in time for the conference

Population structure, multiple paternity, and long-distance transport of spermatozoa in the freshwater mussel Lampsilis cardium (Bivalvia:Unionidae)

Abstract. Freshwater mussels (Bivalvia:Unionidae) are among the most imperiled organisms in North America. Information on the spatial scale of reproduction and population connectivity will better enable mussel conservation programs to sustain long-term population viability, particularly restocking and recovery programs. Here we used genetic methods to characterize population structure, dispersal potential, and reproductive strategies in the freshwater mussel Lampsilis cardium from Twin Creek and Big Darby Creek (Ohio, USA). We genotyped adults and individual glochidia at 12 microsatellite loci to assess local population structure relative to within-population patterns of relatedness and parentage. Local populations within watersheds were weakly structured, and within-population estimates of relatedness identified probable full- and half-siblings several kilometers apart. Parent–offspring comparisons provided evidence of multiple paternity in single broods and identified the likely father of 3 glochidia from 1 females brood 16.2 km upstream of the mother, indicating that long-distance transport of spermatozoa can promote population connectivity within watersheds. Given that lampsilines and other unionoids exhibit similar reproductive strategies, it is possible that other species are capable of long-distance fertilization. If so, fertilization in populations of many freshwater mussels might not be limited by local density of breeding adults. Therefore, the prospects for recovery of imperiled freshwater mussels might be better than what is now expected.

Knowledge discovery in biological data sets using a hybrid Bayes classifier/evolutionary algorithm

A key element of bioinformatics research is the extraction of meaningful information from large experimental data sets. Various approaches, including statistical and graph theoretical methods, data mining, and computational pattern recognition, have been applied to this task with varying degrees of success. We have previously shown that a genetic algorithm coupled with a k-nearest-neighbors classifier performs well in extracting information about protein-water binding from X-ray crystallographic protein structure data. Using a novel classifier based on the Bayes discriminant function, we present a hybrid algorithm that employs feature selection and extraction to isolate salient features from large biological data sets. The effectiveness of this algorithm is demonstrated on various biological and medical data sets.

Systematic Differences in Electropherogram Peak Heights Reported by Different Versions of the GeneScan Software

DNA profiling using STRs on the 310 and 3100 Genetic Analyzers routinely generates electropherograms that are analyzed with the GeneScan software available from the instruments manufacturer, Applied Biosystems. Users have been able to choose from three different smoothing options that have been known to result in significant differences in the peak heights that are reported. Improvements in the underlying algorithm of the most recent version of the software also result in significant and somewhat predictable differences in peak height values. Laboratories that have performed validation studies using older versions of GeneScan should either reanalyze the data generated in those validation studies with the newest version of the software or otherwise take into consideration the systematically higher peak height values obtained as they begin following the recommendation of the manufacturer and use the new algorithm.

PocketMol: a molecular visualization tool for the Pocket PC

Molecular visualization programs are available on many platforms. They allow a user to visualize and manipulate molecular structures. PocketMol provides the same functionality on a Pocket PC handheld computer. Using standard protein data bank (pdb) files, the user can move, rotate, and scale a protein to explore its structure and function. The user can choose from a standard backbone view or a simplified view using only alpha carbon atoms. PocketMolGX uses the Microsoft Game API to provide fast animation that is quite smooth. PocketMol is designed as an aid for those wishing to explore or demonstrate protein structures without the availability of a full-size computer.

A Proposed Statistical Protocol for the Analysis of Metabolic Toxicological Data Derived from NMR Spectroscopy

Nuclear magnetic resonance (NMR) spectroscopy is a non-invasive method of acquiring a metabolic profile from biofluids. This metabolic information may provide keys to the early detection of exposure to a toxin. A typical NMR toxicology data set has low sample size and high dimensionality. Thus, traditional pattern recognition techniques are not always feasible. In this paper, we evaluate several common alternatives for isolating these biomarkers. The fold test, unpaired t-test, and paired t-test were performed on an NMR-derived toxicological data set and results were compared. The paired t-test method was preferred, due to its ability to attribute statistical significance, to take into consideration consistency of a single subject over a time course, and to mitigate the low sample, high dimensionality problem. We then grouped the resulting statistically salient potential biomarkers based on their significance patterns and compared results to several known metabolites affected by the tested toxin. Based on these results, we present a statistical protocol of sequential t-tests and clustering techniques for identifying putative biomarkers. We then present the results of this protocol applied to a specific real world toxicological data set.

A Data Mining Approach to Predicting Phylum for Microbial Organisms Using Genome-Wide Sequence Data

Genomic sequencing projects are generating vast stores of data that provide opportunities and challenges in data analysis. Investigations of trends in codon usage have proven to be a rich area of study in this field. There are a number of methods for isolating codon usage bias in microbial organisms, each designed to capture a specific aspect of the bias. We posit that each species has evolved under the influence of a unique set of environmental constraints that has governed the shaping of the organisms codon usage. Analysis of codon usage data should, therefore, provide insights into the selection process at work influencing genomic composition. To this end, we describe the large-scale mining of genome-level data from several codon usage bias isolation techniques to determine whether this information can be used to predict the phylum and class to which each organism belongs. Successful prediction is an indication that the forces molding the codon usage of a given phylum/class are indeed distinctive, and that it would be of use in understanding the evolutionary forces involved. Additionally, it supports using this method to aid in, and validate existing taxonomic classification techniques.

Perceived Cost of Auxotrophic Amino Acids in Two Bacterial Species

Amino acid biosynthetic pathways are highly conserved throughout all domains of life. Biosynthesis of amino acid requires the diversion of resources from energy production to amino acid production. The consequent energy-cost of producing an individual amino acid is can be estimated by addingthe amount of ATP expended in production itself to the amount of potential energy lost. Some organisms lack the metabolic pathways required for the synthesis of some or all of their amino acids and must obtain them from their environment or their host organism. The energetic costs associated with this means of obtaining amino acids are largely a matter of speculation at the present time. This study examines the perceived cost of auxotrophic amino acids (amino acids an organism is unable to synthesize) in two bacteria (Bacillus cereus ATCC 10987 andVibrio fischeri ES114). Auxotrophic amino acids in bothorganisms were found to be used preferentially in highlyexpressed genes and are therefore likely to be energetically inexpensive relative to those the organisms are capable of synthesizing themselves. A regression approach was used tocomputationally estimate the perceived costs to the organism.