Extending logic programming with statistics dramatically extends the modelling capabilities of the declarative programming paradigm. Questions as to the most probable success-criteria of some goal are now catered for, - as is statistical parameter learning and maximum likelihood algorithms. Much of the strength of stochastical modelling becomes an integrate part of logic programming.
The price of expressive power as always is in computational complexity. Depending on the actual problem domain at hand it may be possible to address the computational problems with suitable program analysis and transformation techniques, and see how existing and optimized implementations for traditional models can be integrated in an efficient way.
Biological Sequence Analysis
The chosen problem domain for the project is that of biological sequence analysis. Finding properties and structures in biological sequence data, for intance DNA, RNA and proteine structure analysis, is an essential driving force in present-day industrial production and research in all areas concerned with living organisms such as drug design , food production, disease prevention, agriculture genetics research, etc. Analysis of biological sequence data requires considerable computer power as well as sophisticated and efficient programs to do the analysis. Current tools based on Hidden Markov Models (HMM) or Stochastic/Probabilistic Context-Free Grammars (SCFG or PCFG) have limitations concerning expressive power, i.e.: what can be modelled and ultimately extracted by analysis. The present project seeks to improve ease of modelling, accuracy and reliability of sequence analysis by using logic-statistic models that are yet largely untested in bioinformatics, and which have a good potential to become a gateway to new scientific results in biology.
For more information, please contact Henning Christiansen.