Antibiotics have made a huge contribution to the extension of human lifespan. However, antimicrobial resistance (AMR) is spreading rapidly, making existing antibiotics ineffective. At the same time, the high rediscovery rate of antibiotics and limited commercial incentives frustrate antibiotic R&D. The co-occurring problems of AMR and nearly empty discovery pipelines form an impending crisis, often referred to as the ‘silent pandemic’. Most of the antibiotics in clinical use are natural products derived from microorganisms. Large-scale bacterial genome sequencing has revealed that most the natural product structural classes have not yet been fully characterized. Similarly, most metabolites observed in mass-spectrometric data cannot be fully dereplicated. The key question we seek to answer in this project is how to effectively prioritize the millions of unknown biosynthetic gene clusters and metabolite features for the discovery of new antimicrobials through predicting structural and functional features of metabolites from genomic and mass-spectrometric data. A major challenge in current antibiotic drug discovery is that genomic, metabolomic, bioactivity and culturing data is scattered across a wide range of resources and is mostly not publicly available. While sharing proprietary data to train machine learning models is not an option, training models on multiple distributed data sources without sharing the data has been an area of active research in various fields. This recent machine learning methodology is generally referred to as federated learning.
In this position, you will set up a federated learning infrastructure and develop multimodal machine learning methods to predict connections between mass spectra, biosynthetic gene clusters, molecular structures, and biological activities, in collaboration with national and international partners. In addition, there will be plenty of opportunities to collaborate with fellow postdocs and PhD students at the Bioinformatics chair group in Wageningen. You will be part of the KIC PRIORITY consortium that aims to improve the antibiotics discovery pipeline. With several academic and industrial partners involved, your work will form the input for other work packages that include the elicitation, expression, and structural elucidation of novel antibiotic candidates, as well as lead generation and optimization. In the end, the research is projected to lead to a new globally accessible infrastructure and new algorithms for multimodal mining of biosynthetic diversity, as well as novel lead compounds that can be taken up by consortium partners to attain concrete societal impact.

Your duties and responsibilities include:

• Setting up federated learning infrastructure to connect omics and phenotypic datasets across public and private international institutes
• Development of machine learning (including deep learning) algorithms to predict links between gene clusters and metabolites, and to predict antimicrobial activities associated with these
• Collaboration with international partners

• Setting up federated learning infrastructure to connect omics and phenotypic datasets across public and private international institutes
• Development of machine learning (including deep learning) algorithms to predict links between gene clusters and metabolites, and to predict antimicrobial activities associated with these
• Collaboration with international partner