Send to

Choose Destination
Math Biosci. 2006 Nov;204(1):119-31. Epub 2006 Apr 17.

Model identification for DNA sequence-structure relationships.

Author information

Cogent Systems Research Laboratory, Department of Electrical Engineering, University of Washington, Seattle, WA 98195, USA.


We investigate the use of algebraic state-space models for the sequence dependent properties of DNA. By considering the DNA sequence as an input signal, rather than using an all atom physical model, computational efficiency is achieved. A challenge in deriving this type of model is obtaining its structure and estimating its parameters. Here we present two candidate model structures for the sequence dependent structural property Slide and a method of encoding the models so that a recursive least squares algorithm can be applied for parameter estimation. These models are based on the assumption that the value of Slide at a base-step is determined by the surrounding tetranucleotide sequence. The first model takes the four bases individually as inputs and has a median root mean square deviation of 0.90 A. The second model takes the four bases pairwise and has a median root mean square deviation of 0.88 A. These values indicate that the accuracy of these models is within the useful range for structure prediction. Performance is comparable to published predictions of a more physically derived model, at significantly less computational cost.

[Indexed for MEDLINE]

Supplemental Content

Full text links

Icon for Elsevier Science
Loading ...
Support Center