Biclustering is an unsupervised classification technique that plays an increasingly important role in the study of modern biology. This data mining technique has provided answers to several challenges raised by the analysis of biological data and more particularly the analysis of gene expression data. It aims to cluster simultaneously genes and conditions. These unsupervised techniques are based essentially on the assumption that the extraction of the co-expressed genes allows to have co-regulated genes. In addition, the integration of biological information in the search process may induce to the extraction of relevant and non-trivial biclusters. Therefore, this work proposes an evolutionary algorithm based on local search method that relies on biological knowledge. An experimental study is achieved on real microarray datasets to evaluate the performance of the proposed algorithm. The assessment and the comparison are based on statistical and biological criteria. A cross-validation experiment is also used to estimate its accuracy. Promising results are obtained. They demonstrate the importance of the integration of the biological knowledge in the biclustering process to foster the efficiency and to promote the discovery of non-trivial and biologically relevant biclusters.

Graph coloring involves assigning colors to the vertices of a graph such that two vertices linked by an edge receive different colors. Graph coloring problems are general models that are very useful to formulate many relevant applications and, however, are computationally difficult. In this work, a general population-based weight learning framework for solving graph coloring problems is presented. Unlike existing methods for graph coloring that are specific to the considered problem, the presented work targets a generic objective by introducing a unified method that can be applied to different graph coloring problems. This work distinguishes itself by its solving approach that formulates the search of a solution as a continuous weight tensor optimization problem and takes advantage of a gradient descent method computed in parallel on graphics processing units. The proposed approach is also characterized by its general global loss function that can easily be adapted to different graph coloring problems. The usefulness of the proposed approach is demonstrated by applying it to solve two typical graph coloring problems and performing extensive computational studies on popular benchmarks. Improved best-known results (new upper bounds) for the equitable graph coloring problem are reported for several large graphs.

AI-assisted molecular optimization is a very active research field as it is expected to provide the next-generation drugs and molecular materials. An important difficulty is that the properties to be optimized rely on costly evaluations. Machine learning methods are investigated with success to predict these properties, but show generalization issues on less known areas of the chemical space. We propose here a surrogate-based black box optimization method, to tackle jointly the optimization and machine learning problems. It consists in optimizing the expected improvement of the surrogate of a molecular property using an evolutionary algorithm. The surrogate is defined as a Gaussian Process Regression (GPR) model, learned on a relevant area of the search space with respect to the property to be optimized. We show that our approach can successfully optimize a costly property of interest much faster than a purely metaheuristic approach.

The objective of this work is to design a molecular generator capable of exploring known as well as unfamiliar areas of the chemical space. Our method must be flexible to adapt to very different problems. Therefore, it has to be able to work with or without the influence of prior data and knowledge. Moreover, regardless of the success, it should be as interpretable as possible to allow for diagnosis and improvement. We propose here a new open source generation method using an evolutionary algorithm to sequentially build molecular graphs. It is independent of starting data and can generate totally unseen compounds. To be able to search a large part of the chemical space, we define an original set of 7 generic mutations close to the atomic level. Our method achieves excellent performances and even records on the QED, penalised logP, SAscore, CLscore as well as the set of goal-directed functions defined in GuacaMol. To demonstrate its flexibility, we tackle a very different objective issued from the organic molecular materials domain. We show that EvoMol can generate sets of optimised molecules having high energy HOMO or low energy LUMO, starting only from methane. We can also set constraints on a synthesizability score and structural features. Finally, the interpretability of EvoMol allows for the visualisation of its exploration process as a chemically relevant tree.

We present frequent pattern-based search (FPBS) that combines data mining and optimization. FPBS is a general-purpose method that unifies data mining and optimization within the population-based search framework. The method emphasizes the relevance of a modular- and component-based approach, making it applicable to optimization problems by instantiating the underlying components. To illustrate its potential for solving difficult combinatorial optimization problems, we apply the method to the well-known and challenging quadratic assignment problem. We show the computational results and comparisons on the hardest QAPLIB benchmark instances. This work reinforces the recent trend toward closer cooperations between the optimization methods and machine learning or data mining techniques.

The extraction of knowledge from large biological data is among the main challenges of bioinformatics. Several data mining techniques have been proposed to extract data; in this work, we focus on biclustering which has grown considerably in recent years. Biclustering aims to extract a set of genes with similar behavior under a condition set. In this paper, we propose an evolutionary biclustering algorithm and we analyze its performance by varying its genetic components. Hence, several versions of the evolutionary biclustering algorithm are introduced. Further, an experimental study is achieved on two real microarray datasets and the results are compared to other state-of-the-art biclustering algorithms. This thorough study allows to retain the best combination of operators among the various experienced choices.

This paper presents an improved probability learning based local search algorithm for the well-known graph coloring problem. The algorithm iterates through three distinct phases: a starting coloring generation phase based on a probability matrix, a heuristic coloring improvement phase and a learning based probability updating phase. The method maintains a dynamically updated probability matrix which specifies the chance for a vertex to belong to each color group. To explore the specific feature of the graph coloring problem where color groups are interchangeable and to avoid the difficulty posed by symmetric solutions, a group matching procedure is used to find the group-to-group correspondence between a starting coloring and its improved coloring. Additionally, by considering the optimization phase as a black box, we adopt the popular tabu search coloring procedure for the coloring improvement phase. We show extensive computational results on the well-known DIMACS benchmark instances and comparisons with state-of-the-art coloring algorithms. (c) 2018 Elsevier B.V. All rights reserved.

A challenging task in systems biology is to decipher cell regulation mechanisms. By comparing networks observed in two different situations, the differential network analysis approach enables to highlight interaction differences that reveal specific cellular responses. The aim of our work is to study the role of natural anti-sense transcription on cellular regulation mechanisms. Our proposal is to build and compare networks obtained from two different sets of actors: the "usual" sense actors on one hand and the sense and anti-sense actors on the other hand. Our study only considers the most significant interactions, called an Extended Core Network; therefore our differential analysis identifies important interactions that are impacted by anti-sense transcription. This paper first introduces our inference method of an Extended Core Network; this method is inspired by C3NET, but whereas C3NET only computes one interaction per gene, we propose to consider the most significant interactions for each gene. Secondly, we define the differential network analysis of two extended core networks inferred with and without anti-sense actors. On a local view, this analysis relies on change motifs that describe which genes have their most important interactions modified when the anti-sense transcripts are considered; they are called AS-impacted genes. Then from a more global view, we consider how the relationships between these AS-impacted genes are rewired in the network with anti-sense actors. Our analysis is performed by computing Steiner trees that represent minimal subnetworks connecting the ASimpacted genes. We show that the visualisation of these results help the biologists to identify interesting parts of the networks.

As a usual model for a variety of practical applications, the maximum diversity problem (MDP) is computational challenging. In this paper, we present an opposition-based memetic algorithm (OBMA) for solving MDP, which integrates the concept of opposition-based learning (OBL) into the well-known memetic search framework. OBMA explores both candidate solutions and their opposite solutions during its initialization and evolution processes. Combined with a powerful local optimization procedure and a rank-based quality-and-distance pool updating strategy, OBMA establishes a suitable balance between exploration and exploitation of its search process. Computational results on 80 popular MDP benchmark instances show that the proposed algorithm matches the best-known solutions for most of instances, and finds improved best solutions (new lower bounds) for 22 instances. We provide experimental evidences to highlight the beneficial effect of OBL for solving MDP.

This paper presents a hybrid approach called frequent pattern based search that combines data mining and optimization. The proposed method uses a data mining procedure to mine frequent patterns from a set of high-quality solutions collected from previous search, and the mined frequent patterns are then employed to build starting solutions that are improved by an optimization procedure. After presenting the general approach and its composing ingredients, we illustrate its application to solve the well-known and challenging quadratic assignment problem. Computational results on the 21 hardest benchmark instances show that the proposed approach competes favorably with state-of-the-art algorithms both in terms of solution quality and computing time.

Un des problemes cles en bioinformatique est de comprendre les me-canismes de regulation au sein d’une cellule. Notre travail concerne l’etude desreseaux de genes chez le pommier, avec la particularite d’y integrer les acteursencore mal connus que sont les ARN anti-sens. Pour explorer l’impact des trans-crits anti-sens, nous proposons ici la comparaison des deux reseaux obtenus parune methode d’inference tres conservative. Nous pouvons ainsi etudier les in-teractions directes entre les genes qui sont modifiees si l’on fait intervenir lestranscrits anti-sens dans la methode d’inference. Un ensemble de motifs caracte-ristiques autour de ces modifications permet de reveler des ensembles d’acteurssens et anti-sens interessants.

Grouping problems aim to partition a set of items into multiple mutually disjoint subsets according to some specific criterion and constraints. Grouping problems cover a large class of important combinatorial optimization problems that are generally computationally difficult. In this paper, we propose a general solution approach for grouping problems, i.e., reinforcement learning based local search (RLS), which combines reinforcement learning techniques with descent-based local search. The viability of the proposed approach is verified on a well-known representative grouping problem (graph coloring) where a very simple descent-based coloring algorithm is applied. Experimental studies on popular DIMACS and COLOR02 benchmark graphs indicate that RLS achieves competitive performances compared to a number of well-known coloring algorithms.

Several transcriptomic studies have shown the widespread existence of anti-sense transcription in cell. Anti-sense RNAs may be important actors in transcriptional control, especially in stress response processes. The aim of our work is to study gene networks, with the particularity to integrate in the process anti-sense transcripts. In this paper, we first present a method that highlights the importance of taking into account anti-sense data into functional enrichment analysis. Secondly, we propose the differential analysis of gene networks built with and without anti-sense actors in order to discover interesting change motifs that involve the anti-sense transcripts. For more reliability, our network comparison only studies the conservative causal part of a network, inferred by the C3NET method. Our work is realized on transcriptomic data from apple fruit.

Cognitive maps are a graphical knowledge representation model that describes influences between concepts, each influence being quantified by a value. Most cognitive map models use values the semantics of which is not formally defined. This paper introduces the probabilistic cognitive maps, a new cognitive map model where the influence values are assumed to be probabilities. We formally define this model and redefine the propagated influence, an operation that computes the global influence between two concepts in the map, to be in accordance with this semantics. To prove the soundness of our model, we propose a method to represent any probabilistic cognitive map as a Bayesian network.

Cognitive maps are a knowledge representation model that describes as a graph influences between concepts. Each influence is quantified by a value. The values are generally not formally defined. In this paper, we introduce a new cognitive map model, the probabilistic cognitive maps. In such maps, the values of the influences are interpreted as probability values. We define formally the semantics of this model. We also provide an operation to compute the global influence of a concept on any other one, called the probabilistic propagated influence. To show that our model is valid, we propose a procedure to represent a probabilistic cognitive map as a Bayesian network.

Microarrays represent a new technology for measuring expression levels of several genes under various biological conditions generating multiple data. These data can be analyzed by using biclustering method which aims to extract a maximum number of genes and conditions presenting a similar behavior. This paper proposes a new evolutionary approach to obtain maximal high-quality biclusters of highly-correlated genes. The performance of the proposed algorithm is assessed on synthetic gene expression data. Experimental results show that our algorithm competes favorably with several state-of-the-art biclustering algorithms.