Pedro Barahona

PSICO: Solving Protein Structures with Constraint Programming and Optimization

In this paper we propose PSICO (Processing Structural Information with Constraint programming and Optimisation) as a constraint-based approach to determining protein structures compatible with distance constraints obtained from Nuclear Magnetic Resonance (NMR) data. We compare the performance of our proposed algorithm with DYANA (“Dynamics algorithm for NMR applications”) an existing commercial application based on simulated annealing. On a test case with experimental data on the dimeric protein Desulforedoxin, the method proposed here supplied similar results in less than 10 minutes compared to approximately 10 hours of computation time for DYANA. Although the quality of results can still be improved, this shows that CP technology can greatly reduce computation time, a major advantage because structural NMR technique generally demands multiple runs of structural computation.

Modelling Digital Circuits Problems with Set Constraints

A number of diagnostic and optimisation problems in Electronics Computer Aided Design have usually been handled either by specific tools or by mapping them into a general problem solver (e.g. a propositional Boolean SAT tool). This approach, however, requires models with substantial duplication of digital circuits. In Constraint Logic Programming, the use of extra values in the digital signals (other than the usual 0/1) was proposed to reflect their dependency on some faulty gate. In this paper we present an extension of this modelling approach, using set variables to denote dependency of the signals on sets of faults, to model different circuits problems. We then show the importance of propagating constraints on sets cardinality, by comparing Cardinal, a set constraint solver that we implemented, with a simpler version that propagates these constraints similarly to Conjunto, a widely available set constraint solver. Results show speed ups of Cardinal of about two orders of magnitude, on a set of diagnostic problems.

Applying Constraint Programming to Protein Structure Determination

In this paper, we propose a constraint-based approach to determining protein structures compatible with distance constraints obtained from Nuclear Magnetic Resonance (NMR) data. We compare the performance of our proposed algorithm with DYANA (“Dynamics algorithm for NMR applications” [1]) an existing commercial application based on simulated annealing. For our test case, computation time for DYANA was more than six hours, whereas the method we propose produced similar results in 8 minutes, so we show that the application of Constraint Programming (CP) technology can greatly reduce computation time. This is a major advantage because this NMR technique generally demands multiple runs of structural computation.

Global Hull Consistency with Local Search for Continuous Constraint Solving

This paper addresses constraint solving over continuous domains in the context of decision making, and discusses the trade-off between precision in the definition of the solution space and the computational effort required. In alternative to local consistency, which is usually maintained in handling continuous constraints, we discuss maintaining global hull-consistency. Experimental results show that this may be an appropriate choice, achieving acceptable precision with relatively low computational cost. The approach relies on efficient algorithms and the best results are obtained with the integration of a local search procedure within interval constraint propagation.

Computerising a guideline for the management of diabetes

This paper reports an experience of computerising a clinical guideline for the management of non-insulin-dependent diabetes mellitus (NIDDM). The guideline, designed by the European NIDDM Policy Group is being used in a National Programme for Diabetes supported by the Portuguese Ministry of Health, who is keen to supporting its widespread use by general practitioners, namely in computerised form. The paper presents the main characteristics of the prototype that was implemented within the European project Prestige, and was developed according to the Prestige Protocol Model. The model is briefly described, together with the generic architecture that supports it. Then the main design decisions of the prototype are explained, regarding the modelling of a general practitioner workflow during a typical consultation and the user interface, two key issues for obtaining acceptance from the users. The limitations of the system are discussed and a number of directions are outlined in order to circumvent such limitations, and broaden the scope of applicability of the system.

Paper: A causal and temporal reasoning model and its use in drug therapy applications

Superficial knowledge about drug effects and interactions may provide clinicians with only a limited support for the elaboration of therapy plans. Deeper knowledge of the mechanisms through which drugs produce their effects, together with their temporal constraints, should be modelled to predict the effects and interactions of their joint administration. The present paper describes a method for modelling such deep medical knowledge, together with its causal and temporal reasoning capabilities, and compares it with classical approaches to temporal and causal reasoning, namely in the context of drug treatment applications. This method extends a previous causal functional model by allowing the representation of quantitative knowledge, the explicit representation of time intervals, and a temporal reasoning technique, Simulation by Interval Constraining. The method is illustrated by a number of examples of basic drug metabolic mechanisms; and its future use in the development of decision support systems for complex drug therapy applications is discussed.

Constraint Programming in Structural Bioinformatics

Bioinformatics aims at applying computer science methods to the wealth of data collected in a variety of experiments in life sciences (e.g. cell and molecular biology, biochemistry, medicine, etc.) in order to help analysing such data and eliciting new knowledge from it. In addition to string processing bioinformatics is often identified with machine learning used for mining the large banks of bio-data available in electronic format, namely in a number of web servers. Nevertheless, there are opportunities of applying other computational techniques in some bioinformatics applications. In this paper, we report the application of constraint programming to address two structural bioinformatics problems, protein structure prediction and protein interaction (docking). The efficient application of constraint programming requires innovative modelling of these problems, as well as the development of advanced propagation techniques (e.g. global reasoning and propagation), which were adopted in Chemera, a system that is currently used to support biochemists in their research.

Applying constraint programming to rigid body Protein Docking

In this paper we show how Constraint Programming (CP) techniques can improve the efficiency and applicability of grid-based algorithms for optimising surface contact between complex solids. We use BiGGER [1] (Bimolecular complex Generation with Global Evaluation and Ranking) to illustrate the method as applied to modelling protein interactions, an important effort in current bioinformatics. BiGGER prunes the search space by maintaining bounds consistency on interval constraints that model the requirement for the shapes to be in contact but not overlapping, and by using a branch and bound approach to search the models with the best fit. This CP approach gives BiGGER some efficiency advantages over popular protein docking methods that use Fourier transforms to match protein structures. We also present an efficient algorithm to actively impose a broad range of constraints or combinations of constraints on distances between points of the two structures to dock, which allows the use of experimental data to increase the effectiveness and speed of modelling protein interactions and which cannot be done as efficiently in Fourier transform methods. This shows that constraint programming provides a different approach to protein docking (and fitting of shapes in general) that increases the scope of application while improving efficiency.

Constraint reasoning in deep biomedical models

Objective:: Deep biomedical models are often expressed by means of differential equations. Despite their expressive power, they are difficult to reason about and make decisions, given their non-linearity and the important effects that the uncertainty on data may cause. The objective of this work is to propose a constraint reasoning framework to support safe decisions based on deep biomedical models. Method:: The methods used in our approach include the generic constraint propagation techniques for reducing the bounds of uncertainty of the numerical variables complemented with new constraint reasoning techniques that we developed to handle differential equations. Results:: The results of our approach are illustrated in biomedical models for the diagnosis of diabetes, tuning of drug design and epidemiology where it was a valuable decision-supporting tool notwithstanding the uncertainty on data. Conclusion:: The main conclusion that follows from the results is that, in biomedical decision support, constraint reasoning may be a worthwhile alternative to traditional simulation methods, especially when safe decisions are required.

A Framework for Optimal Correction of Inconsistent Linear Constraints

The problem of inconsistency between constraints often arises in practice as the result, among others, of the complexity of real models or due to unrealistic requirements and preferences. To overcome such inconsistency two major actions may be taken: removal of constraints or changes in the coefficients of the model. This last approach, that can be generically described as “model correction” is the problem we address in this paper in the context of linear constraints over the reals. The correction of the right hand side alone, which is very close to a fuzzy constraints approach, was one of the first proposals to deal with inconsistency, as it may be mapped into a linear problem. The correction of both the matrix of coefficients and the right hand side introduces non linearity in the constraints. The degree of difficulty in solving the problem of the optimal correction depends on the objective function, whose purpose is to measure the closeness between the original and corrected model. Contrary to other norms, that provide corrections with quite rigid patterns, the optimization of the important Frobenius norm was still an open problem. We have analyzed the problem using the KKT conditions and derived necessary and sufficient conditions which enabled us to unequivocally characterize local optima, in terms of the solution of the Total Least Squares and the set of active constraints. These conditions justify a set of pruning rules, which proved, in preliminary experimental results, quite successful in a tree search procedure for determining the global minimizer.