Mots-Clés

deep learning multi-omics explanations integration

Description

Interpretable multi-omics deep learning framework in reproductive biology

Recently, deep learning methods involving artificial neural networks (ANNs) with multiple layers have become popular to perform classification and regression tasks involving large amounts of data. They have been successfully applied in many domains such as image recognition, robotics speech recognition, life sciences, etc. ANNs can handle complex and noisy data, and can also model nonlinear relationships thanks to the layer-wise architecture of non-linear processing units.

Multi-omics approaches allow us to understand how information flow one molecular level to the another. They can provide reliable and novel information. The main obstacle of current omics studies is the observation or study of the system through a single level of biological complexity. 

The objective of this project will be to  study, apply, develop and improve ANN based multi-omics approaches. The aim is to build a framework where ANNs based on different architectures can be combined into the final model. We will explore how we can extend the explanation method to understand predictions from multiple ANNs. Student will have access to the large amount of omics data related to the female reproductive system available in the team.

Expected skills

Background in computers science and programming skills (Python), having a good knowledge of machine learning algorithms.

Contact: misbah(dot)razzaq(at)INRAE(dot)fr