InCeTo: Simplify Complexity and Increase Certainty in Toxicology through Systematic Identification and Understanding of Molecular Key Principles
This project is a UFZ TB CITE PhD Colleg project adressing the topic: Deriving principles for chemical impact assessment.
The project frames 4 independent but connected PhD projects across 4 departments:
- AUC: Mechanistic description of uptake and distribution
of selected ions to cells and organisms, supervised by Dr. Nadin Ulrich
- BIOTOX: Development of a zebrafish embryo TKTD model based on time- and dose-resolved omics data for selected ionic compounds, supervised by Dr. Wibke Busch
- MOLSYB: Unravelling the mode of action (MoA) of ionic chemicals and cross-species approaches, supervised by Dr. Kristin Schubert
- PhD IV - BIOINF: Advanced computational investigation of ionic compounds in a toxicological setting, supervised by Dr. Jana Schor, details see below
The assessment and understanding of the impact of chemicals on human and environmental health are vital for guiding actions in terms of regulations and management. The behavior and impact of chemicals on organisms depend on chemical structures, which determine their properties and interactions with biomolecules. For neutral chemicals, these properties, and interactions are well studied and understood. The calculation and assessment of mixture toxicity of such compounds are straightforward with established principles such as concentration addition. However, only a tiny fraction of chemicals relevant to human and environmental health occur in their neutral form, e.g., about 80% of chemicals measured in European fresh waters occur as ions. Applying and advancing the existing models to ionic chemicals is challenging, e.g., due to the lack of available labeled information on the charge and toxicity of chemicals. A systematic mechanistic understanding of toxicokinetic (TK) and toxicodynamic (TD) processes from exposure to effect is hardly available yet.
With this project, we aim to gain a mechanistic and principle understanding of physicochemical parameters and biological key events and a deep understanding of molecular responses, which determine the fate and activity of ionic compounds. Therefore, we will combine expertise on existing models to predict and understand the chemical distribution, chemical toxicity/activity, and biological action. Furthermore, using state-of-the-art chemical analytics, bio-analytics, modeling, machine learning on chemical structures, multi-omics data analysis, integration, and comparative genomics, a principle understanding of these processes for ionic compounds across species, based on TK and TD processes, should be gained. Finally, the project results should lead to cross-disciplinary chemical assessment with a solid computational component to move from descriptive observations to a systematic understanding of relevant processes for toxicity across species. Furthermore, we want to strengthen the interdisciplinary usage of data in toxicology. Therefore, we provide a showcase and teach how data- and model-driven hypotheses generation leads to systematic process understanding in toxicology.
PhD IV BIOINF - Advanced computational investigation of ionic compounds in a toxicological setting
This Ph.D. position is part of the InCeTo TB Colleg Increasing certainty in toxicology. The project covers three further Ph.D. projects that focus on the detailed understanding of ionic compounds regarding their uptake, distribution, effects, and toxicity in vitro. This Ph.D. is a Data Science position focused on investigating toxicity on a large scale – ideally the chemical universe.
The idea is to enhance the existing deepFPlearn (dfpl) approach, which associates a chemical's molecular structure with the cellular effect on the gene/pathway level. In dfpl, a deep autoencoder reduces the feature space, and a deep feed-forward neural network classifies the reduced feature set as active or not w.r.t. a particular cellular effect.
Two challenges are obvious, 1) the input format – currently a 2048 bit binary fingerprint that describes the molecular structure of the chemical, and 2) the binary association to the outcome. This project focuses on graph neural networks' potential and regression strategies to address both challenges. Further, we will improve the input data regarding their toxicity associations and evaluate the change in prediction performance when we use this data for training.