1,721,060 research outputs found
Knowledge graphs and explainable artificial intelligence : application to drug repositioning
No full textLe repositionnement de médicaments consiste à trouver de nouvelles utilisations thérapeutiques pour des médicaments existants qui sont déjà approuvés pour traiter d’autres pathologies. Cette approche profite des connaissances déjà existantes sur ces molécules, permettant ainsi un développement plus rapide et moins coûteux par rapport à la création de nouveaux médicaments. Le repositionnement est particulièrement utile pour répondre à des besoins médicaux non satisfaits, comme par exemple pour les maladies rares ou émergentes. Ces dernières années, le développement de graphes de connaissances a permis de concentrer toutes ces informations biomédicales autour du médicament issues de grandes bases de données ou de connaissances. Un graphe de connaissances est une représentation structurée d’informations provenant de différentes sources, qui relie ces informations les unes aux autres par l’utilisation de relations. Cette représentation est particulièrement utile pour mieux comprendre les relations complexes qui structurent nos connaissances sur un médicament. Elle est utilisée de nos jours pour la tâche de repositionnement en particulier. Une façon efficace de repositionner des médicaments à partir de ces graphes est d’utiliser des méthodes d’intelligence artificielle qui prédisent de nouveaux liens entre les objets du graphe. De cette manière, un modèle correctement entraîné sera capable de proposer une nouvelle connexion entre un médicament et une maladie, indiquant une potentielle opportunité de repositionnement. Cette méthodologie présente cependant un gros désavantage : les modèles pour la prédiction de liens fournissent souvent des résultats opaques, qui ne peuvent pas être interprétés par l’utilisateur final des prédictions. Cette thèse propose d’étudier l’utilisation de méthodes d’intelligence artificielle explicables dans le but de repositionner des médicaments à partir de données biomédicales représentées dans des graphes de connaissances. Dans un premier temps, nous analysons l’impact du pré-entraînement sur les modèles de multihop reasoning pour la prédiction de liens. Nous montrons que la construction des représentations des entités du graphe avant l’entraînement du modèle permet une amélioration des performances prédictives, ainsi que de la quantité et la diversité des explications. Dans un second temps, nous étudions comment l’ajout de relations dans un graphe de connaissances affecte les résultats de prédiction de liens. Nous montrons que l’ajout de liens dans trois graphes biomédicaux permet une amélioration des performances prédictives du modèle SQUIRE, et ce sur différents types de relations lien avec le repositionnement de médicaments. Une analyse de l’impact sur l’explicabilité du modèle est aussi menée à la suite de l’ajout de ces relations. Enfin, nous proposons une nouvelle méthodologie pour la tâche de classification de liens dans un graphe de connaissances, basée sur l’utilisation de forêts aléatoires. À partir des informations concernant le voisinage de chaque noeud dans le graphe, nous montrons qu’un modèle de forêts aléatoires est capable de prédire correctement l’existence ou non d’un lien entre deux noeuds. Ces résultats permettent une visualisation des noeuds utilisés pour réaliser la prédiction. Enfin, nous appliquons cette méthode au repositionnement de médicaments pour la sclérose latérale amyotrophique (SLA).Drug repositioning involves finding new therapeutic uses for existing medications that are already approved to treat other conditions. This approach takes advantage of the existing knowledge about these molecules, enabling faster and less costly development compared to creating new drugs. Repositioning is particularly useful for addressing unmet medical needs, such as rare or emerging diseases. In recent years, the development of knowledge graphs has enabled the consolidation of all this biomedical information around drugs, coming from large data sources or knowledge repositories. A knowledge graph is a structured representation of information integrated from different sources, linking these pieces of information together using relationships. This representation is especially useful for understanding the complex relationships that structure knowledge about drugs. Nowadays, it is widely used for the task of drug repositioning. An effective way to reposition drugs using these graphs is to employ artificial intelligence (AI) methods that predict new links between objects in the graph. In this way, a well-trained model can suggest a new connection between a drug and a disease, indicating a potential opportunity for repositioning. However, this methodology has a significant disadvantage : link prediction models often provide opaque results that cannot be easily interpreted by the end users. This thesis proposes to explore the use of explainable AI methods for the purpose of repositioning drugs based on biomedical data represented in knowledge graphs. First, we analyze the impact of pre-training on multihop reasoning models for link prediction. We demonstrate that building representations of the graph entities before model training improves the predictive performance, as well as the quantity and diversity of explanations. Secondly, we examine how the addition of relationships in a knowledge graph affects link prediction results. We show that adding links in three biomedical knowledge graphs improves the predictive performance of the SQUIRE model across different types of relationships related to drug repositioning. An analysis of the impact on model explainability is also conducted, following the addition of these relationships. Finally, we propose a new methodology for the task of link classification in a knowledge graph, based on the use of random forests. Using information about the neighborhood of each node in the graph, we show that a random forest model can accurately predict the existence or absence of a link between two nodes. These results allow for a visualization of the nodes used to make the predictions. Lastly, we apply this method to drug repositioning for amyotrophic lateral sclerosis (ALS)
Improving ICT4D projects with agile software development
No full textICT4D seeks to bridge the digital divide in developing countries. Important requirements of ICT4D projects are a demand-driven approach and participation of the local community. The fact that user collaboration is a principle of Agile software development (Agile), triggers our interest on whether Agile practices can improve ICT4D projects. This paper aims to investigate if and how Agile can contribute to the success of ICT4D projects. In order to achieve this, existing literature was consulted and an interview was held. This paper provides an overview of the critical success factors for ICT4D projects and Agile, as well as of the advantages of Agile. Agile can only work successfully when ICT4D projects are demand-driven, and when both a cultural understanding and trust are built. Notable ways in which Agile can improve ICT4D projects are by facilitating user collaboration, improving team communication, enhancing organizational learning, and by frequently delivering software
Scalability factors in an ICT4D context: A literature review
No full textThis research investigates possible scalability factors that influence an ICT4D project. By performing a literature study on four strands of literature, which include: technical literature (1), development studies (2), technology adoption (3) and ICT4D literature (4), it was found that there are seventeen factors that need to be accounted for in the development process. Furthermore, a general outline of an ICT4D development process is presented and scalability factors are related to phases in this ICT4D process. Future research could focus on validating these factors by using them in a development cycle and determining the precise influence, rather than determining an overall positive or negative influence
Going Beyond Counting First Authors in Author Co-citation Analysis
Full text linkThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
Variations on the Author
Full text link“Variations on the Author” discusses two of Eduardo Coutinho’s recent films (Um Dia na Vida, from 2010, and Últimas Conversas, posthumously released in 2015) and their contribution to the general question of documentary authorship. The director’s filmography is characterized by a consistent yet self-effacing form of authorial self-inscription: Coutinho often features as an interviewer that rather than express opinions propels discourses; an interviewer that is good at listening. This mode of self-inscription characterizes him as an author who is not expressive but who is nonetheless markedly present on the screen. In Um Dia na Vida, however, Coutinho is completely absent form the image, while Últimas Conversas, on the contrary, includes a confessional prologue that moves the director from the margins to the center of his films. This article examines the ways in which these works stand out in the filmography of a director who offers new insights into the notion of cinematic authorship
Appropriate Similarity Measures for Author Cocitation Analysis
Full text linkWe provide a number of new insights into the methodological discussion about author cocitation analysis. We first argue that the use of the Pearson correlation for measuring the similarity between authors’ cocitation profiles is not very satisfactory. We then discuss what kind of similarity measures may be used as an alternative to the Pearson correlation. We consider three similarity measures in particular. One is the well-known cosine. The other two similarity measures have not been used before in the bibliometric literature. Finally, we show by means of an example that our findings have a high practical relevance.information science;Pearson correlation;cosine;similarity measure;author cocitation analysis
Dispelling the Myths Behind First-author Citation Counts
Full text linkWe conducted a full-scale evaluative citation analysis study of scholars in the XML research field to explore just how different from each other author rankings resulting from different citation counting methods actually are, and to demonstrate the capability of emerging data and tools on the Web in supporting more realistic citation counting methods. Our results contest some common arguments for the continued
use of first-author citation counts in the evaluation of scholars, such as high correlations between author rankings by first-author citation counts and other citation
counting methods, and high costs of using more realistic citation counting methods that are not well-supported by the ISI databases. It is argued that increasingly available digital full text research papers make it possible for citation analysis studies to go beyond what the ISI databases have directly supported and to employ more
sophisticated methods
- …
