[Dynamite] telegraph edits

[Ian Holmes]

BTW -- here are my edits, just in case www.biowiki.org goes down while I'm
away.

----

<H2>Telegraph Poster Abstract for ISMB 2000</H2>

----

*Telegraph:* A new dynamic programming template library

_Ian Holmes, Guy Slater, Ewan Birney, Gerald M Rubin._

----

*Dynamic Programming* (DP) is the foundation of a large number of
algorithms in bioinformatics, including protein comparison, gene
prediction and database search using profile hidden Markov models (such as
those found in Pfam). Each of these algorithms aligns sequence(s) to a
*Finite State Machine* (FSM), though the FSM's architecture is different
in each case. The transitions of the FSM have associated "weights"; these
weights determine the scoring scheme of the algorithm (gap penalties,
substitution matrices and so forth). If the weights are normalised, then
the FSM is a probabilistic model and the alignment scores may be treated
as log-likelihoods. This representation is convenient for "training" the
model from data.

The ubiquitous nature of DP in bioinformatics has motivated development of
general purpose libraries for specifying algorithms in terms of the
underlying FSM (e.g. Searls & Murphy, ISMB, 1995; Birney & Durbin, ISMB,
1997). Taking ideas from several of these libraries (most notably
"Dynamite", Birney & Durbin) we have built a new object-oriented library
for DP in bioinformatics, called *Telegraph*.

Telegraph offers the following advantages over existing
libraries:

	* Easily scriptable. The outer interface to the library is written in Perl and integrates smoothly into a Bioperl installation
	* Intuitive. The object model combines the specialised data structure-oriented features of a language like Dynamite, with functional programming-inspired features such as binding of model parameters to user-defined functions
	* Portable to many architectures. The core algorithms have a well-defined ANSI C API
	* Parseable XML file format for models
	* Built-in and integrated support for training algorithms, including Baum-Welch with Dirichlet mixture priors

This poster will illustrate Telegraph with a number of real life
examples.



----

_I think this is probably fine, but please improve it if you can !_

----
Back to the FrontPage
----
*Related pages:* %Unclassified%
----
$LASTMODIFIED:959535748$
$AUTHOR:63.206.105.131$