Category: Uncategorized

  • Seminar on Dec 10th, 2019, 3:00 PM – MLG hosting Dr. Benjamin Haibe-Kains from Princess Margaret Cancer Centre, University Health Network talking about “Dealing with noisy phenotypes to build more robust predictors of drug response in cancer”.

    Title:
        “Dealing with noisy phenotypes to build more robust predictors of drug response in cancer”
    By:
        Dr. Benjamin Haibe-Kains
    When:
        Tuesday 10th December, at 15.00
    Where:
        Université libre de Bruxelles,
        Campus de la Plaine (http://www.ulb.ac.be/campus/plaine/plan.html),
        NO Building, 8th Floor, NO8.08, Rotule room (http://www.ulb.ac.be/campus/plaine/plan-NO.html),
        Boulevard du Triomphe,
        1050 Bruxelles, Belgium.
    Abstract:
    One of the main challenges in precision oncology consists of developing predictors of drug response to select the most beneficial therapy for each individual patient. In this context, preclinical models are crucial to study the association between molecular features of tumor cells and response to chemical perturbations. However, only few predictors have been successfully translated to clinical settings. Such a low success rate is due not only to the complexity of the mechanisms underlying anticancer drug response, but also to the lack of robustness of the predictors developed in preclinical settings (i.e., using patient-derived tumor models like immortalized cancer cell lines). To address this issue we developed PharmacoGx, a computational platform enabling meta-analysis of large-scale drug screenings of in vitro and in vivo model systems, and PharmacoDB (pharmacodb.ca), a web-application enabling quick access to a large compendium of pharmacogenomics datasets. In this presentation I will show how we used our new platforms to develop univariate and multivariate predictors of drug response in the presence of high levels of phenotypic noise to generate candidate companion diagnostics that can be considered for testing in future clinical trials. 

    Biography:
    Dr. Benjamin Haibe-Kains is a Senior Scientist at the Princess Margaret Cancer Centre (PM), University Health Network, and Associate Professor in the Medical Biophysics department of the University of Toronto. Dr. Haibe-Kains earned his PhD in Bioinformatics at the Université Libre de Bruxelles (Belgium). Supported by a Fulbright Award, he did his postdoctoral fellowship at the Dana-Farber Cancer Institute and Harvard School of Public Health (USA).
    He started his own laboratory at the Institut de Recherches Cliniques de Montréal (Canada) and moved to PM in November 2013. Dr. Haibe-Kains’ research focuses on the integration of high-throughput data from various sources to simultaneously analyze multiple facets of carcinogenesis. In particular, Dr. Haibe-Kains and his team are analyzing radiological and (pharmaco)genomic datasets to develop new prognostic and predictive models and to discover new therapeutic strategies with the aim to significantly improve disease management. Dr. Haibe-Kains’ main scientific contributions include several prognostic gene signatures in breast cancer, subtype classification models for ovarian and breast cancers, genomic predictors of drug response in cancer cell lines, and radiomic prognostic models in head-and-neck cancers.

  • MLG co-organizes AI Synergies

    MLG co-organizes AI Synergies

    This year’s Benelux AI (BNAIC)and Machine Learning (Benelearn) conferences will be organised in Brussels. The conferences are hosted by the ULB Machine Learning Group and the AI lab of the VUB. More information can bound on the website http://ai-synergies.be/

  • MLG and (IB)2 in VRT Nieuws and L’echo

    MLG and (IB)2 in VRT Nieuws and L’echo

    The Flemish news site VRT nieuws and the French news paper L’Echo are reporting on the excellent work of the oligogenic team. Congratulations!

    L’Echo: À Bruxelles, les maladies rares sont pistées grâce à l’intelligence artificielle
    https://www.lecho.be/r/t/1/id/10133875

    VRT Nieuws : VUB en ULB ontwikkelen AI-methode om genetische oorzaken zeldzame ziekten te identificeren | VRT NWS
    https://www.vrt.be/vrtnws/nl/2019/06/05/vub-en-ulb-ontwikkelen-ai-methode-om-genetische-oorzaken-zeldzam.html/

  • Call for papers BNAIC/BENELRARN

    Call for papers BNAIC/BENELRARN

    The call for papers for the Benelux conferences on Artificial Intelligence are launched. More information can be found on http://bnaic19.brussels and http://benelearn19.brussels

  • Looking for a PhD!

    Looking for a PhD!

    We are looking for a PhD student for the FWO funded project DELICIOS. The research will be performed with the groups of Prof. Pieter Simoens (UG – IDlab), Prof. Jo Pierson (VUB – SMIT)  and Prof. Tom Lenaerts (VUB/ULB – AI lab/MLG).  This works builds on the expertise of these groups in collective intelligence, (evolutionary) game theory, human trust, behavioral intelligence and AI.  Your work will consist of building agent-based models, performing a series of behavioral experiments, as defined in the research context below, report on the scientific insights gathered from these models and experiments.

    To apply see : https://www.imec-int.com/nl/work-at-imec/job-opportunities/phd-position-in-behavioral-informatics-and-ai

    Research context: In this age of ubiquitous digital interconnectivity, we may envisage that humans will increasingly delegate their social, economic or data-related transactions to an autonomous agent, for reasons of convenience or complexity. Although the scientific knowledge to create such systems appears to be available, this transformation does not appear to become commonplace soon, except maybe the use of basic digital assistants.

    We aim to explore if this is due to the lack of knowledge about human trust and acceptance of artificial autonomous delegates that make decisions in their place or even how these delegates should be designed. We will study these questions using computational agent-based models that are validated in a series of behavioral experiments defined around the public goods game. We investigate when and how the autonomous agent may evolve from observer, over decision support to a delegate with full autonomy in decision-making. In later phases of the research, we will investigate networking effects between these agents and study the collective behavior of such agent-based systems.

    The research is conducted in the scope of a multi-disciplinary project. You will collaborate with two other researchers. One other student will investigate how a VR/AR representation of the agent influences trust. The other student – from the domain of social sciences – will check all the technology-oriented research against socio-technology acceptance theories. The results of this fundamental research will allow us to explore important questions related to the intelligence and interface of the envisioned agents and lay the foundation for new types of online markets that bring autonomous agents into real-world applications.

  • New MLG publication at the International Conference on Machine Learning (ICML) 2019

    New MLG publication at the International Conference on Machine Learning (ICML) 2019

    Congratulations to Axel Abels for having his paper entitled “Dynamic Weights in Multi-Objective Deep Reinforcement Learning”, accepted for presentation and for publication in the conference proceedings at ICML 2019!!

  • AI4Belgium Launched today

    AI4Belgium Launched today

    Today a coalition of academic, public and private sector specialists launched the Belgian strategy for AI, originally called AI4Belgium. All information concerning the recommendations from this panel can be found on the coalition website.

    Summary of recommendations of the AI 4 Belgium Coalition:

    We have world-class assets that need to be nurtured and developed. And with the right level of ambition and thoughtful implementation, we can change our society for the better. We structure our recommendations in five chapters. We start with skills, putting people first, and with a responsible way of sharing data. Technology should be at our service, not the other way around. The next three chapters focus on technology adoption, innovation and better public service. In conclusion, we set out a few implementation principles, such as the need for overall ambition.

    Set up a new learning deal – Technology and AI are transforming society and our job market. We currently lack both the capacity and tools to support this transition and our schools are not preparing the next generations for the 21stcentury. This is the reason why we propose a new learning deal; a universal skills building program for adults and more digital – as well as human – skills for our youth.

    Develop a responsible data strategy – Trust is the cornerstone of any transformation. We believe in the need for a robust and up-to-date legal framework, ethical principles and more transparency. Moreover, data is the energy that will fuel the fourth industrial revolution. But data often remains inaccessible. We need to build a data ecosystem that facilitates more responsible data-sharing with reinforced open data policies, more collaborations and a platform with well-structured tools and approaches.

    Support private sector AI adoption – It can be hard for companies, particularly SMEs, to start working with AI. It can be perceived as complex; companies might lack the internal resources and the iterative approach can be too costly. Hence, we propose to demystify AI through a lighthouse approach (training programs, large-scale events and social-impact projects). Secondly, we believe in more collaboration and accessibility to AI through a national AI hub. Lastly, we need to facilitate experimentation.

    Innovate and radiate – We have world-class researchers, but our research is not at scale. Also, we have yet to develop, attract and retain enough AI talent. Lastly, it is hard for innovative start-up companies to grow beyond the early stages. Hence, we propose to position Belgium as Europe’s AI lab through sandboxes and large-scale collaboration within academia, leveraging Belgian transposition of the GDPR. Next, we recommend creating more AI-related training programs, more focus on practical applications and more selective migration. Lastly, we suggest supporting the growth of our AI companies through an investment fund and by differentiating our expertise.

    Improve public service and boost the ecosystem –Too few public organisations are currently experimenting with AI. Firstly, we propose that public institutions rethink their own roles and evolve towards a platform approach. Secondly, we need to give public institutions the tools to experiment; such as a rolling fund and more innovation-friendly procurement. Lastly, we recommend creating a Chief Digital Officer role to organise internal transformations and launch large-scale transversal projects.

    A few principles to ensure a sustainable implementation: ensuring continued trust from the public, a European approach, collaboration between all stakeholders, a grass-roots/community-led approach, focus on specific areas (such as healthcare/life sciences) and, lastly, daring to beambitious and audacious. This will require an investment of at least EUR 1 billion by 2030.