Mots-Clés

Apprentissage profond deep learning génomique genomics épigénétique epigenetics métabolomique metabolomics données clinique clinical data pytorch keras tensor flow neural networks réseaux de neurones

Description

Missions

The Centre national de médecine de précision des diabètes (PreciDIAB) seeks a talented post-doc to develop artificial intelligence tools – particularly Deep Learning – to better understand, predict, prevent, and target diabetes and related metabolic diseases. The successful applicant will be part of the UMR1283/8199 led by Professor Philippe Froguel and the European Genomic Institute for Diabetes, located at the CHU of Lille, internationally recognised for their research on the topic. They will join the team “Metabolic functional (epi)genomics and their abnormalities in type 2 diabetes and related disorders”, an interdisciplinary group of experts in genomics and computational biology led by Amélie Bonnefond to strenghten a novel major axis of research at the institute centred on artificial intelligence.

Activities

AI is revolutionising life sciences and medicine research. Deep Learning has been used in diabetes for some ten years for the prediction, detection and classification of the disease, as well as for glycaemic control and the diagnosis of complications. However, the learning cohorts used are quite small, and the algorithms used do not make the best use of recent developments, particularly in patient stratification and personalised, precision medicine. The successful applicant will contribute to the development of the next generation of AI solutions applied to diabetes and related disorders. They will be involved in all aspects of the design, implementation, and deployment of AI solutions, including:

• Preparation and analysis of heterogeneous datasets, including clinical, genetics, (epi)genomics, transcriptomics and metabolomics data;

• Design, programming, and deployment of composite neural networks, based on several architecture and trained on heterogeneous data;

• Assessment and validation of the models’ predictions in collaboration with biological and clinical experts;

• Use of interpretable AI methods to extract meaningful features from the models and derive new biological insights, such as molecular networks underpinning the metabolic disorders.

The postdoctoral fellow be responsible for the preparation, interpretation, and dissemination of results, including writing research articles and presenting in conferences.

Skills

We are looking for an expert in artificial intelligence already trained in Deep Learning approaches, and able to hit the ground running. A good knowledge of molecular and cellular biology or physiology is appreciated but not mandatory.

• AI-related PhD in computer science, mathematics, physics, or bioinformatics;

• Excellent knowledge of Deep Learning, the underlying concepts, the different architectures, including Convolutional networks, Auto-encoders, Transformers and Large Language Models;

• Proficiency in Python programming, with good knowledge of the Tensor/Keras or the Pytorch universe; knowledge of R programming appreciated;

• Knowledge of working with GPU, and the CUDA framework;

• Good knowledge of a variety of machine learning approaches;

• Good experience of Unix/Linux and working with distributed computing infrastructures;

• Mastery of the English language (written and spoken);

• Curiosity, rigour, and desire to work in a highly collaborative environment.

Work Context

The successful applicant will join the team “Metabolic functional (epi)genomics and their abnormalities in type 2 diabetes and related disorders”, an interdisciplinary and very collaborative group of experts in genomics and computational biology recognised worldwide. They will have access to unique cohorts of patients providing clinical, genomic, and functional genomic data. The research is supported by technological platforms and supporting staff, such as NGS, metabolomics, bioinformatics, and biostatistics.

On top of an excellent intellectual environment the institute provides the infrastructure required to carry out the project, with direct access to state-of-the-art platforms including large computing clusters with thousands of CPUs and hundreds of GPUs, together with petabytes of storage.

PreciDIAB https://www.precidiab.org/

UMR1283/8199 http://www.good.cnrs.fr/

EGID https://egid.fr/