Doctoral Candidates

 
​Project title: Machine learning-guided multiparametric optimisation of cytochrome P450 monooxygenase

The aim of this project is to simultaneously improve the activity and stability of cytochrome P450 BM3, minimizing trade-off effects and enabling a smoother transition to larger-scale application. To enable this, Lisa will develop a workflow applying high-throughput screening to generate experimental data on the enzyme's sequence and functional properties, which will be used to iteratively train a predictive machine learning model.

Background fields: structural biology, molecular biology

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.

Project title: ML-guided evolution of peroxygenases for selective C-H oxyfunctionalisation reactions

Toni's project focuses on combining machine learning and directed evolution to engineer unspecific peroxygenases (UPOs) for selective oxyfunctionalization reactions. The project aims to integrate computational approaches with experimental enzyme engineering to explore UPO sequence–function relationships and develop improved biocatalysts.

Background fields: molecular biotechnology, enzyme engineering, directed evolution, protein engineering, machine learning, biocatalysis

Project title: Multiparametric engineering of alkene dioxygenase

Sebastián's project is focused on the characterization and improvement of non-heme iron dependent Alkene Dioxigenases, in a multiparametric approach. He will use both rational design and random mutagenesis to accomplish his goals of improving the fitness of our collection of alkene dioxigenases.

Background fields: molecular biology, microbiology, bioinformatics

Project title: Engineering conformational dynamics and mechanism in oxygenases

The project focuses on the computationally guided design of oxygenases through the engineering of enzyme conformational dynamics and catalytic mechanism. The project aims to understand how changes in protein sequence, including mutations at residues distant from the active site can reshape the conformational ensamble of oxygenases and thereby improve activity and selectivity. In particular, Shortest Path Map (SPM) analyses are used to investigate communication networks within the enzyme and identify pathways connecting distal residues to the active site.

Background fields: bioinformatics, machine-learning, drug discovery workflows, cheminformatics

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.

Project title: Kinetic stability data collection for predictive bioprocess modelling of CYP enzymes

The main objectives of this project are to understand how dissolved oxygen affects enzyme stability in a bioprocess-realistic environment, as well as to develop a hybrid model to predict enzyme stability across different reactors and conditions. The project will involve a scale-down approach, with a focus on gathering relevant stability data for P450. The collected data will then be used for the development of a hybrid model utilizing active learning and machine learning. The initial work will focus on isolated enzymes, with the ambition of expanding the model to whole-cell systems.

Background fields: molecular biotechnology, fermentation, biocatalysis

Project title: Process intensification of cytochrome P450 reactions in batch and continuous reactors

This project focuses on intensifying cytochrome P450-catalysed reactions in batch and continuous-flow reactors. Controlled bioreactor experiments and reaction engineering will be used to investigate how substrate supply, and the delivery of oxygen or hydrogen peroxide influence reaction efficiency and productivity. In collaboration with WP1, engineered P450 variants will be evaluated under process-relevant conditions, while the resulting kinetic and process data will contribute to modelling and machine-learning activities in WP3. Selected reactions will also be investigated in pressurized continuous-flow systems to improve oxygen availability. By identifying key process bottlenecks and developing scale-up criteria, the project aims to connect enzyme engineering with robust and scalable biocatalytic production.

Background fields: biotechnology, biochemistry, biocatalysis, bioprocess engineering, continuous-flow biocatalysis

Project title: Process development and optimization of dioxygenase catalysed biotransformation

The project is process development and optimisation of dioxygenase-catalysed biotransformation based on the conversion of isoeugenol to vanillin, focusing on biocatalyst operational stability under bioreactor conditions. The research will combine experimental biocatalyst development, kinetic characterization, mathematical modelling, and in silico process intensification and validation.

Background fields: enzyme bioprocess, reaction engineering,  process intensification, machine learning

Project title: Continuous flow oxyfunctionalisations of allylic substrates catalysed by UPOs

The aim of this project is to develop efficient and reusable unspecific peroxygenase (UPO) biocatalysts for selective oxyfunctionalization. Nektaria will combine protein engineering and process engineering to improve UPO activity, selectivity, stability, and reusability for allylic oxyfunctionalization. Diverse UPOs and allylic substrates will be screened to identify promising enzyme–substrate combinations, while enzyme immobilization, continuous-flow operation, and controlled H₂O₂ supply will be explored to enhance overall catalytic and process performance, enabling more efficient and sustainable oxyfunctionalization chemistry.

Background fields: biotechnology, biochemistry, biocatalysis, protein engineering, and bioprocess engineering 

Project title: Machine learning-based tools for multiparametric enzyme optimisation

Daniela's project focuses on developing machine learning-based tools for multiparametric enzyme optimisation. Building on protein mutational datasets generated in-house for a range of enzymes, as well as public protein databases, she will explore and compare state-of-the-art data representation techniques to build next-generation data-driven models for engineering properties such as enzyme activity, selectivity, stability and solubility.

Background fields: computational chemistry, bioinformatics, molecular dynamics

Project title: Development of machine learned coarse grained force fields for computational enzyme design

Filippo works on machine-learned force-fields for enzyme systems — neural network models that learn from reference data the potential energy surface governing how these proteins move. He is particularly interested in coarse-grained models, which enable the simulation of long-timescale dynamics. He is also interested in explainability and interpretability methods, which make machine-learning models transparent rather than black boxes: understanding what the network has learned about the underlying physics and why it produces a given prediction, both to trust the simulations and to extract mechanistic insight into how enzymes work.

Background fields: physics, machine learning

Project title: Development of ML-based tools for analysis and engineering protein dynamics

Tillman’s project focuses on developing machine-learning tools to analyse and engineer protein dynamics. These movements and conformational changes play a central role in how enzymes function. The project aims to move beyond static protein structures by developing efficient methods to predict, understand, and deliberately modify dynamic behaviour. He will also develop AI-assisted workflows that support the progression from researching a target protein and computationally screening designs to prioritizing candidates for experimental validation. The broader goal is to establish protein dynamics as a practical component of protein design, enabling useful enzyme variants to be identified with fewer experiments and less computation.

Background fields: machine learning, bioinformatics, scientific software engineering, agentic workflows

Project title: EVB- and AI-guided Engineering of Unspecific Peroxygenases (UPOs) for Selective Benzylic Hydroxylation

Biya's project combines computational methods and machine learning to engineer unspecific peroxygenases (UPOs) for selective benzylic C–H hydroxylation, a reaction that competes with off-target epoxidation and overoxidation. Towards this, she will build calibrated EVB/QM/MM models of the enzyme's reaction mechanism to understand how its structure governs this selectivity, and use these insights to propose targeted mutations. The collected simulation data will then be used to develop an uncertainty-aware active learning model that prioritises the enzyme variants most worth testing, rather than screening them all. The project also extends into process engineering, modelling how the enzyme performs under real operating conditions to support its future use at an industrial scale. This combination of mechanistic simulation and data-driven prediction is intended to identify UPO variants and operating conditions that improve selectivity and productivity, with fewer experiments than conventional screening would require.

Background fields: theoretical and computational chemistry, molecular dynamics simulation

Project title: Hybrid models for data-driven bioprocess optimization

Athina's project focuses on developing novel machine learning methods for the design and optimization of bioprocesses in biocatalysis. The objective is to develop an algorithmic toolbox for bioprocess optimization, scale-down process development, control and transfer learning across different processes. The optimization aims to maximize process outcome and minimize experimental resources. Regarding scale-up and scale-down, variable mismatches arise between scales, leading to different effects on the process. For control, the goal is to achieve real-time control under uncertainty, accounting for the nonlinear and dynamic nature of the system. The model will be applied to UPO, CYP and DO production processes, using hybrid models as the modeling method.

Background fields: bioprocess engineering, hybrid modeling, machine learning