Contribution à la modélisation et à l'explication de l'identification de cibles thérapeuriques en utilisant des modèles d'intelligence artificielle hybride basés sur la modélisation de connaissances et des systèmes multiagents.

Contributing to the Modeling and Explanation of therapeutic Target Identification using Hybrid Artificial Intelligence based on Knowledge Modeling, Agent-based modeling and Explainable Artificial intelligence approaches

• Auteur / Autrice : Simon Stephan
• Direction : Stéphane Galland, Ouassila Narsis labbani
• Type : Projet de thèse
• Discipline(s) : Informatique
• Date : Inscription en doctorat le 15/10/2021
• Etablissement(s) : Bourgogne Franche-Comté
• Ecole(s) doctorale(s) : École doctorale Sciences pour l'ingénieur et microtechniques (Besançon ; 1991-....)
• Partenaire(s) de recherche : Laboratoire : Connaissance et Intelligence Artificielle Distribuées
• Partenaire(s) de recherche : établissement de préparation : Université de technologie de Belfort-Montbéliard (1999-....)

Mots clés

Mots clés libres

Ingénierie de connaissance

Modélisation ontologique

Systèmes multiagents

Optimisation par metaheuristique orientée agents

Ciblage thérapeutique

Knowledge Engineering

Ontological Modelings

Multiagent systems

Agent-based metaheuristic optimization

Therapeutic Targeting

Résumé

The problem of therapeutic target identification is facing different problems: 1) Creating a knowledge representation of the concepts (from biology) that are involved in the target identification problem. 2) Creating an algorithm that is exploring the space of solutions for target identification and finding the relevant ones with accuracy and efficiency. 3) Explain the found solutions to the human experts in such a way the computation results could be understood and validated by them. These three key problems will be considered and studied with artificial intelligence approaches. More specifically, the PhD candidate will have to contribute to the definition of an ontology in collaboration with the other researchers involved in the project from CIAD laboratory and Oncodesign company. Regarding the exploration algorithm, we have pre-selected the scientific field of multigent systems [9] and metaheuristic optimization based on multiagent systems. Indeed, collective natural bio-systems like ant and bee colonies, flocks of birds and swarms, as well as systems of cells and molecules, are composed of multiple bio-entities residing in the physical environment and engaged in complex collective and organized behaviors, interactions and processes according to the laws of nature. There is a certain level of abstraction at which the behavior of such systems can be modeled as distributed computational processes resulting from the interaction of artificial computational entities. Thus, we would expect distributed computing to have a lot of potential for the practical application of therapeutic target identification. For example, ant colony optimization (ACO) [1] is inspired by the collective behavior of colonies of natural ants when they explore the environment searching for food. During their search process, ants secrete pheromone on their way back to their anthill. Other ants of the colony sense the pheromone and are attracted to the marked paths; the more pheromone that is deposited on a path, the more attractive that path becomes. The pheromone is volatile and disappears over time. Evaporation erases the pheromone on longer paths as well as on less interesting paths. Shorter paths are refreshed more quickly, thus having the chance of being more frequently explored. Intuitively, ants will converge towards the most efficient path solution due to the fact that it gets the strongest concentration of pheromone. Other approaches of agent-based metaheuristic optimization could be found in literature based on Boids [2] or modeling based other insect behavior (bees, etc.) [3]. The PhD student will explore these different approaches and propose a novel algorithm for solving the therapeutic target identification problem. Regarding the third major problem, we would like to explore the adaptation of an artificial intelligence explanation (XAI) model. Explaining the reasoning and the outcomes of complex computer programs has received considerable attention since the 1990's when research works on explainable expert systems were disseminated. Nowadays, with the pervasive applications of machine learning, the need of explaining the reasoning of Artificial Intelligence is considered a top priority. In 2017, the European Parliament recommended AI systems to follow the principle of transparency; systems should be able to justify their decisions in a way that is understandable to humans [6]. In April 2019, the European Union High-Level Expert Group on AI presented the Ethics Guidelines for Trustworthy AI [7]. This report highlighted transparency as a key property of trustworthy AI. In the same vein, recent works in the literature highlighted explainability as one of the cornerstones for building trustworthy, responsible, and acceptable AI systems. Consequently, the sub-domain research of eXplainable Artificial Intelligence (XAI) gained momentum both in academia and industry. XAI research aims at explaining the outcomes of AI-driven systems [8] since, in the absence of a proper explanation, the human user will come up with an explanation that might be flawed or erroneous. In turn, this will degrade the user's acceptability. Intelligent agents have been established as a suitable technique for implementing these systems requiring autonomous high-level control and decision-making in complex AI systems [9]. In this context of heterogeneous human-agent systems, recent works in goal-driven XAI aims at ensuring mutual understandability, improving acceptability, and enhancing human-agent collaboration capabilities. In the context of this PhD thesis, the foundations of XAI for therapeutic target identification will be explored and defined in order to explain the results provided by the agent-based metaheuristic optimization model that is at the heart of this PhD thesis.

Le problème de l'identification des cibles thérapeutiques est confronté à différents problèmes : 1) Créer une représentation des connaissances des concepts (issus de la biologie) qui sont impliqués dans le problème de l'identification des cibles. 2) Créer un algorithme qui explore l'espace des solutions pour l'identification des cibles et découvre efficacement les solutions pertinentes. 3) Expliquer les solutions trouvées aux experts humains de manière à ce que les résultats calculés puissent être compris et validés par eux. Ces trois problèmes clés seront considérés et étudiés avec des approches d'intelligence artificielle. Plus spécifiquement, le doctorant devra contribuer à la définition d'une ontologie en collaboration avec les autres chercheurs impliqués dans le projet du laboratoire CIAD et de la société Oncodesign. Concernant l'algorithme d'exploration, nous avons présélectionné le domaine scientifique des systèmes multiagents et de l'optimisation métaheuristique basée sur les systèmes multiagents. En effet, les bio-systèmes naturels collectifs comme les colonies de fourmis et d'abeilles, les volées d'oiseaux d'essaims, ainsi que les systèmes de cellules et de molécules, sont composés de multiples entités biologiques résidant dans un environnement physique commun et engagées dans des comportements, des interactions et des processus collectifs complexes. Il existe un certain niveau d'abstraction auquel le comportement de ces systèmes peut être modélisés comme des processus informatiques distribués résultant des interactions entre entités informatiques artificielles. Par conséquent, le calcul distribué possède un grand potentiel pour l'application pratique dans l'identification de cibles thérapeutiques. Le doctorant explorera différentes approches de modélisation relatives à l'optimisation metaheuristique orientée agents et proposera une approche et modèle permettant de répondre à la problématique de ciblage thérapeutiques. Concernant le troisième problème scientifique, nous souhaitons explorer l'adaptation d'un modèle d'intelligence artificielle explicable (XAI).