[bull-ia] PhD Proposal


PhD Proposal: Development of machine learning algorithms for the identification of biomarkers of neurotoxicity.


Context : H2020 Project NeuroDeRisk

Advisors: Engelbert Mephu Nguifo (engelbert.mephu_nguifo@uca.fr)

Vincent Barra (vincent.barra@isima.fr)


Location : LIMOS, UMR 6158 CNRS (https://limos.fr/)

Clermont-Auvergne University

Scientific Campus, 1 rue de la Chébarde



Dates: September 2019-August 2022


There is still a lack of complete understanding of the complex pathways and mechanisms leading to central and peripheral drug-induced neurotoxicity. However, these adverse effects are to a considerable extent (up to 25%) responsible for failure in clinical trials, affecting volunteers or patients administered with experimental drugs, and also, to some extent, albeit less than 1%, for hospitalizations caused by hidden neurotoxic adverse effects of marketed drugs.


The ambition of NeuroDeRisk[i] H2020 project is to bundle the scientific expertise of experimental and theoretical scientists and to collaborate with computer scientists to address the challenge of preclinical prediction of neurotoxicity using a fully integrated approach: by linking unique expertise for building in vitro and in vivo models with in silicoprediction tools, the project will establish a novel and validated toolbox for preclinical prediction of neurotoxicity in humans.


The PhD subject is centered on the development of new machine learning strategies to optimize the identification of predictive and alerting biomarkers of neurotoxicity. The objectives here will be:

–        to classify subjects (with respect to neurotoxicity based on data collected by previous extensive behavioral and biochemical signatures of neurotoxicity in vivo, in vitro and ex vivo) ;

–        to compute low dimensional manifolds on which subjects lie, and be able to explore  the trends and directions of all possible neurotoxic effects in the derived space ;

–        to be able to early predict biomarkers of neurotoxicity, based on only a subset of data or on new features derived from the original data.

Both classical and innovative machine learning algorithms will be explored.


Having selected the most promising biomarker candidates from in vivo, in vitro, and ex vivo studies, in silico prediction of biomarkers using a suitably modified version of a predictive platform will be performed.  Additional bioinformatics tools will be used in biomarker validation and prioritization by other partners of the European project.



[i] NeuroDeRisk (Neurotoxicity De-Risking in Preclinical Drug Discovery) is a H2020 European Project gathering 19 partners from 13 countries.