Scope and Focus

This Laboratory's research involves the application of computers to solving problems in chemistry and biochemistry. We have several projects involving computer-aided analysis of data obtained from chemical instrumentation. We develop new methods and apply new and old methods to learn more about the behavior of chemical and biochemical systems. Chemical applications associated with these projects range over a wide area, including understanding and improving monitoring batch reactions for chemical production, standoff detection of gases by hyperspectral imaging, geochemical prospecting, and genomic and chemoinformatic analyses.

Undergraduates pursuing any of these projects may work with graduate students, post-doctoral workers and also will work directly with Prof. Brown. Students interested in any of these projects should have some prior computer experience with either Linux, Macintosh or Windows. Some exposure to computer programming and experience with chemical instrumentation is helpful, but not essential.

Data and Model Fusion for Multivariate Classification and Calibration

We are investigating ways in which data from the same samples measured on different instruments can be combined for use in creating better mathematical models for classification or calibration. Opportunities exist to take data on a variety of chemical instrumentation and to analyze the data separately and in combination.

Improving the Robustness and Transfer of Multivariate Calibrations

This project involves the development of new methods for improving quantitative measurements of components in mixtures. Most aspects of this project involve the analysis of spectral data, but some opportunities are available for measurement of near-infrared, visible and ultraviolet spectra of mixtures on several different spectrometers. The spectral data set is then used to build multivariate chemical calibrations for each data set collected. The focus of the project is on using mathematical methods implemented on computers for building quantitative calibrations for components of the mixtures that are robust to changes in the instruments over time or to use of the calibration with data from another spectrometer.

Novel Methods for Improved Classification of High-Dimensional Data Sets
We are developing new ways of grouping chemical data together to discover hidden relationships among the samples and uncover the most informative measurements for discovering those relationships. We are combining ideas from research on  decision trees and various forms of Bayesian networks to create new classifiers and new means of discovery of information in large data sets. These new tools are being applied to a wide range of data sets in chemistry and biochemistry.

©2007 University of Delaware

Site Created by S.D. Brown URL of this page: http://www.udel.edu/Links/chemo_UG_res.htm

Last revision: 28 August 2007