No description available.

Ces dernières années, les thérapies ciblées pour le traitement du cancer, ont été introduites. Cependant, un médicament fonctionnant chez un patient peut ne pas fonctionner chez un autre. Pour éviter l'administration de traitements inefficaces, des méthodes capables de prédire les patients qui répondront à un médicament donné doivent être mises au point.Il n'est actuellement pas possible de prédire l'efficacité de la grande majorité des médicaments anticancéreux. L’apprentissage automatique (AA) est un outil particulièrement prometteur pour la médecine personnalisée. L’AA est un champ d’étude de l'intelligence artificielle ; elle concerne la mise au point et l'application d'algorithmes informatiques qui s'améliorent avec l'expérience. Dans ce cas, l'algorithme d’AA apprendra à faire la distinction entre les tumeurs sensibles et résistantes en fonction de plusieurs gènes au lieu d'un seul gène. Cette étude se concentre sur l'application de différentes approches de l’AA pour prédire la réponse à des médicaments anticancéreux des tumeurs et générer des modèles précis, biologiquement pertinentes et faciles à expliquer. / In recent years, targeted drugs for the treatment of cancer have been introduced. However, a drug that works in one patient may not work in another patient. To avoid the administration of ineffective treatments, methods that predict which patients will respond to a particular drug must be developed.Unfortunately, it is not currently possible to predict the effectiveness of most anticancer drugs. Machine learning (ML) is a particularly promising approach for personalized medicine. ML is a form of artificial intelligence; it concerns the development and application of computer algorithms that improve with experience. In this case, ML algorithm will learn to distinguish between sensitive and non-sensitive tumours based on multiple genes instead of a single gene. Our study focuses on applying different approaches of ML to predict drug response of tumours to anticancer drugs and generate models which have good accuracy, as well as are biologically relevant and easy to be explained.

During the past few decades, the landscape of available therapeutic interventions for cancer treatment has widely expanded, boosted mainly by immunotherapy progress and the precision oncology paradigm. The extensive use of pre-clinical models in cancer research has led to the discovery of new effective treatment options for patients. Despite the notable advancements, some cancer types have found minor benefits from the use of precision-oncology interventions. Characterized by a heterogeneous molecular landscape, bladder cancer is one of the most frequent cancer types in which standard-of- care treatments involve surgical operations accompanied by broad-spectrum chemotherapy. My research stems from the need for precision oncology interventions in bladder cancer and specifically focuses on the development of computational frameworks to guide the discovery of new therapeutic opportunities. This work first introduces the exploration of possible therapeutic interventions in 9p21.3 depleted bladder tumors through the analysis of an in-house large High-Content Drug Screening that tested 2,349 compounds. By combining cell count changes and morphological quantitative features extracted from fluorescence images, we nominate cytarabine as a putative candidate eliciting specific cytotoxic effects in an engineered 9p21.3 depleted bladder cancer model compared to an isogenic wild-type clone. Focusing on the development of computational methodologies to nominate robust context-specific vulnerabilities, I further describe PRODE (PROtein interactions informed Differential Essentiality), an analytical workflow that integrates protein-protein interaction data and Loss of Function screening data. I extensively tested PRODE against the most commonly used and alternative methodologies and demonstrated its superior performance when classifying reference essential and context-essential genes collected from experimental and literature sources. Furthermore, we applied PRODE to a real case scenario, seeking essential genes selectively in the context of HER2+ Breast Cancer tumors. Finally, I report the computational analyses performed on Patient-Derived Organoids (PDOs) established from a bladder cancer cohort. PDOs are demonstrated as informative models when assessing the therapeutic sensitivity of patients to drugs. Overall, this research highlights novel precision-oncology applications by ad-hoc computational analyses that address key open technical and biological challenges in the field of bladder cancer and beyond.

No description available.