Activity fingerprints in DNA based on a structural analysis of sequence information.

The function of a DNA sequence is commonly predicted by measuring its nucleotide
similarity to known functional sets. However, the use of structural properties to
identify patterns within families is justified by the discovery that many very different
sequences have similar structural properties. The aim of this thesis is to develop tools
that detect any unusual structural characteristics of a particular sequence or that
identify DNA structure-activity fingerprints common to a set.
This work uses the Octamer Database to describe DNA. The database's contents are
split into two categories: those parameters that describe minimum energy structure and
those that measure flexibility. Information from both of these categories has been
combined to describe structural tendencies, offering an alternative measure of sequence
similarity.
A structural DNA profile gives a graphical illustration of how a parameter from the
Octamer Database varies across either a single sequence's length or across a set of
sequences. Profile Manager is an application that has been developed to automate
single sequence profile generation and is used to study the A-tract phenomenon. The
use of profiles to explore patterns in flexibility across a set of pre-aligned promoters is
then investigated with interesting transitions in decreasing twist flexibility discovered.
Multiple sequence queries are harder to solve than those of single sequences, due to the
inherent need for the sequences to be aligned. It is only under rare circumstances that
sequences are pre-aligned by an experimentally determined position. More commonly
a multiple alignment must be generated. An extended, structure-based, hidden Markov
model technique that successfully generates structural alignment~ is presented. Its.
application is tested on four DNA protein binding site datasets with comparisons made
to the traditional sequence method. Structural alignments of two out of the four
datasets were comparable in performance to sequence with useful insights into
underlying structural mechanisms.