Mots-Clés

Machine learning on multi-omics data, feature selection, support mechanistic models with machine learning

Description

Objectives

The objective of this ESR project is to train young researchers with strong programming and/or data scientist background to tackle machine and deep learning problems in the field of systems biology and multi-omics. As an example, development of mechanistic models will be supported by machine learning, for instance by pinpointing inconsistencies in the models with the data and by identifying links between variables not captured in the models. Traditional machine learning approaches include unsupervised learning and clustering to reason about the network links and role in genome-scale models. Tree based methods (random forest, decision tree) provide a natural way to analyse the contribution of different features in a model. More recent sequence-based learning methods like recurrent neural networks are able to use contextual information. In these models there is no need for advance feature selection, since it is encoded in the weights of the network. Although the interpretability of these models is limited, in practice they perform well if sufficiently large amount of data is available. Word/sentence embedding methods are state of the art ways to represent sequential data to deep neural networks in analogue natural language processing tasks. In this project sequence embedding techniques are investigated as an input layer of neural networks. The machine learning models are utilized in dynamics modelling and process modelling tasks.

Required Education Level

Master’s degree in computer science or equivalent degrees ideally with a strong background in data science, machine learning, deep learning no later than September 2021. You should NOT have any kind of PhD degree. Previous research experience (which must be no longer than 4 years) although appreciated, is not mandatory.

Skills / Qualifications

• Computer science, data science, machine learning, deep learning background
• Provable software development skills in one of common programming languages (e.g. Java, C++) Python, Tensorflow, PyTorch experience is favoured
• Willingness to learn the necessary biology background in order to work with biological data properly
• Educational background and previous research experience relevant for the chosen position
• Networking and good communications skills (writing and presentation skills)
• Willingness to travel abroad for the purpose of research, training and dissemination