Ole Torp Lassen

Inference with constrained hidden markov models in prism

A Hidden Markov Model (HMM) is a common statistical model which is widely used for analysis of biological sequence data and other sequential phenomena. In the present paper we show how HMMs can be extended with side-constraints and present constraint solving techniques for efficient inference. Defining HMMs with side-constraints in Constraint Logic Programming have advantages in terms of more compact expression and pruning opportunities during inference. We present a PRISM-based framework for extending HMMs with side-constraints and show how well-known constraints such as cardinality and all different are integrated. We experimentally validate our approach on the biologically motivated problem of global pairwise alignment.

Bayesian Annotation Networks for Complex Sequence Analysis

Probabilistic models that associate annotations to sequential data are widely used in computational biology and a range of other applications. Models integrating with logic programs provide, furthermore, for sophistication and generality, at the cost of potentially very high computational complexity. A methodology is proposed for modularization of such models into sub-models, each representing a particular interpretation of the input data to be analysed. Their composition forms, in a natural way, a Bayesian network, and we show how standard methods for prediction and training can be adapted for such composite models in an iterative way, obtaining reasonable complexity results. Our methodology can be implemented using the probabilistic-logic PRISM system, developed by Sato et al, in a way that allows for practical applications.

Preprocessing for Optimization of Probabilistic-Logic Models for Sequence Analysis

A class of probabilistic-logic models is considered, which increases the expressibility from HMMs and SCFGs regular and context-free languages to, in principle, Turing complete languages. In general, such models are computationally far too complex for direct use, so optimization by pruning and approximation are needed. The first steps are taken towards a methodology for optimizing such models by approximations using auxiliary models for preprocessing or splitting them into submodels. Evaluation of such approximating models is challenging as authoritative test data may be sparse. On the other hand, the original complex models may be used for generating artificial evaluation data by efficient sampling, which can be used in the evaluation, although it does not constitute a foolproof test procedure. These models and evaluation processes are illustrated in the PRISM system developed by other authors, and we discuss their applicability and limitations.

Biosequence Analysis in PRISM

In this work, we consider probabilistic models that can infer biological information solely from biological sequences such as DNA. Traditionally, computational models for biological sequence analysis have been implemented in a wide variety of procedural and object oriented programming languages [1]. Models implemented using stochastic logic programming (SLP [2,3,4]) instead, may draw upon the benefits of increased expressive power, conciseness and compositionality. It does, however, pose a big challenge to design efficient SLP models.

A Declarative Pipeline Language for Complex Data Analysis

We introduce BANpipe – a logic-based scripting language designed to model complex compositions of time consuming analyses. Its declarative semantics is described together with alternative operational semantics facilitating goal directed execution, parallel execution, change propagation and type checking. A portable implementation is provided, which supports expressing complex pipelines that may integrate different Prolog systems and provide automatic management of files.