Terry Clark - Research Interests
This page is way under construction

Molecular Modeling of Nucleosomes (with Levi Pierce, University of Kansas)
   Nucleosomes consist of a 147-base-pair span of DNA tightly wrapped around a histone-protein octamer. In 1.7 turns the DNA bends sharply at helical-period positions where a significant bias for certain dinucleotides has been found experimentally. We use molecular modeling to investigate the mechanistic properties behind the observed dinucleotides biases in work geared to elucidate DNA sequence effects. Modeling of free-DNA provides a considerable foundation for these studies. However, specialized studies are needed for the nucleosome due to non-uniform super-helical nature of the DNA and close interactions with the histone core.

Genome Analysis (with John Goldsmith and L. Ridgway Scott, University of Chicago)
   Sequence similarity is an indispensible counterpart to electronic protein and DNA sequences, used in various applications ranging from identifying mRNA products to producing and annotating genome sequences. At the base of sequence similarity analysis is a dynamic-programming algorithm that finds the optimal alignment between two sequences based on a score matrix. Alignment statistics developed independently by Bill Pearson and Samuel Karlin attach a probability to the score in order to determine the biological significance of the similarity.
   We are developing non-alignment based methods as a basis to detect features and characterize genome sequence. In contrast to sequence alignment, our methods use the notion of a lexicon of nucleotide words. The lexicon is inductively acquired in an optimization procedure using expectation maximization. Essentially, words are frequently occurring strings in the sequence, which may be of arbitrary length. A new application, we have been successful in using our multi-gram model to characterize chromosome properties (manuscript in progress). This and future work involve isolating significant motifs giving rise to the characterization.

Parallel Computing
   After at least 20 years of efforts in parallel programming methodology, parallel computing is as difficult as ever with language support an open question actively engaged. The implications for robust programming are on the rise with developments in multi-core processors, configurable FPGAs, wide-spread use of commodity clusters, and the likely appearance of commodity boards with hundreds of processors.
   We have developed a parallel programming paradigm, the Planguages, with programming constructs supporting explicit parallelism. This model has been successfully used in developing industrial-strength parallel applications and in education. Currently, we are interested in extending the Planguages to include shared-memory constructs and run-time system alternatives to the message-passing approach underlying the current compiler implementations.

Genomics Data Management
   The object-oriented Genomics Unified Schema (GUS) developed at the University of Pennsylvania provides an extensive genomics oriented database schema with an object-oriented interface and centralized management framework for genomics data management. We developed XMLGUS as an alternative to the one-plugin-per-input-data-type approach common with GUS. XMLGUS extends the GUS object layer to generate an interface automatically for a standard GUS XML format, resolving keys constraints as needed. This system further unifies the GUS data interface, provides support for dynamically extending the GUS schema with new data types, and generally facilitates interoperation. The new release reported here automates generation of interface components based on the database schema description. XMLGUS 2.0 is available at flora.ittc.ku.edu with examples and notes. Future work includes optimizing the interface to reduce overhead, such as is encountered in inputting and annotating chromosomes or large numbers of sequence reads. In ongoing work, we have also experimented with an approach to extract data from the using the same XML description database.