See all projects

This is an alphabetic list of all the projects contained in our website:

  • Cytotoxicity · In this project, cytotoxicity prediction, one of the earliest handles in drug discovery, is investigated using a deep learning approach trained on a large and highly consistent in-house data set from the Leibniz-Forschungsinstitut für Molekulare Pharmakologie (FMP). While neural networks are often described as a black box, we try to overcome the lack of interpretability. Thus, a Deep Taylor Decomposition method is investigated to identify substructures that may be responsible for the cytotoxic effects.
  • DeepLearningVS · DeeplearningVS is a project which aims to study a novel rescoring method based on deep learning (DL) techniques to enhance the accuracy of docking results, and boost the structure-based virtual screening outcome.
  • ECFT VS-Pipeline · Finding the optimal docking pipeline to identify novel ligands for a target of interest is challenging. Our project aims to develop a data-driven pipeline to find new hits for energy-coupling factor (ECF) transporters through the optimization of various structure-based virtual screening workflows.
  • Historical Virtual Control Groups · Historical Virtual Control Groups: one step forward into the future of animal testing in toxicology. Our goal is to reduce the number of animals used in experiments. Starting with an exceptional dataset provided by members of the eTRANSAFE consortium we start the journey into the future of animal testing via derivation and incorporation of virtual control groups in animal testing approaches and thus enabling a 3R strategy.
  • KinfragML · In this project, we work with the protein kinase family, since they are involved in many diseases which make them an interesting drug target. The aim is to generate novel ligands as potential drug candidates for kinases using generative deep learning. The model includes structural information on the kinase binding pocket, since it is well studied and well conserved across protein kinases. We tackle the problem of ligand generation in a fragment-based approach by including domain-knowledge to increase kinase affinity and synthesizability in general.
  • KinFragLib · The KinFragLib project allows to explore and extend the chemical space of kinase inhibitors using data-driven fragmentation and recombination, built on available structural kinome data from the KLIFS database for over 2,500 kinase complexes. The computational fragmentation method splits known non-covalent kinase inhibitors into fragments with respect to their 3D proximity to six predefined subpockets relevant for binding.
  • Kinodata-3D · Machine learning - and especially deep learning - models require large datasets for training. As such datasets, especially those containing protein-ligand-complex information - are more rare in the drug design landscape, we assess the use of in silico structural docking data for machine learning. To this end, we perform template docking using the OpenEye software on a large kinase activity dataset (kinodata) following the complex generation pipeline developed in kinoml.
  • KinoML · In this project, we aim to combine structure-enabled machine learning and alchemical free energy calculations to develop a predictive quantitative model to rapidly assess kinase inhibitor affinity and selectivity, design ligands with desired selectivity profiles and assess the impact of clinical point mutations on inhibitor binding.
  • KiSSim · Kinases are important and well studied drug targets for cancer and inflammatory diseases. Due to the highly conserved structure of kinases, especially at the ATP binding site, the main challenge when developing kinase inhibitors is achieving selectivity, which requires a comprehensive understanding of kinase similarity. In our KiSSim (Kinase Structural Similarity) project, we developed a subpocket-focused kinase fingerprint to investigate kinome-wide pocket similarity.
  • KnowTox · KnowTox is a novel pipeline that combines three different in silico toxicology approaches to allow for confident prediction of potentially toxic effects of novel chemical entities, i.e. machine learning models, alerts for toxic substructures and computational support for read-across.
  • Maxsmi · Deep learning requires lots of data which in the case of physico- chemical and bioactivity remains scarce. Here, we exploit that one compound can be represented by various SMILES strings as means of data augmentation and we explore several augmentation techniques. The best strategies lead to the Maxsmi models, the models that maximize the performance in SMILES augmentation. These models are trained on four data sets, including experimental solubility, lipophilicity, and bioactivity measurements, and are available for prediction on novel compounds.
  • MIAME · In this collaboration with Bayer and the Chodera Lab, we aim to advance and apply KinoML to address pharmaceutically relevant drug design challenges. Special emphasis is put on the effect of point mutations on binding affinity and how these can be exploited to expand the indications of already approved drugs and to guide molecular design.
  • OpenCADD · OpenCADD is a Python library for structural cheminformatics.
  • PLIPify (WIP) · Protein-Ligand Interaction Frequencies across Multiple Structures.
  • PROTECT · In this project, we aim to explore the potential of transformer models to optimize molecular properties and/or generate new molecules with desired non-toxic properties. Unlike currently available machine and deep learning methods, self supervised learning models, e.g., the transformer architecture, provide generalizability by being pre-trained on large unsupervised dataset then fine-tuned on small downstream datasets. The transformer model is highly resourceful with the ability to perform molecular property prediction, optimization, and/or generation.
  • Ratar · How to probe and validate a potential target remains one of the key questions in basic research in life sciences. In the DFG-project Ratar (Read-Across the Targetome), we will use binding site similarity to predict off-targets and to extrapolate compound information from one target to another. This similarity-based knowledge transfer can suggest tool compounds (chemical probes) and off-targets for proteins of interest using the ever-growing amount of available target and compound data.
  • T²F-Flex · T²F-Flex extends the functionality of T²F-Pharm and offers pharmacophore modeling from MD simulations of apo structures, i.e. empty binding sites, to represent the dynamic nature of the structures. Interaction hotspots on the grid are calculated per snapshot (subset of the trajectory) and subsequently clustered and summarized into a single pharmacophore model. This method allows to generate pharmacophores in the absence of ligand or interaction information and provides a feature-based description of protein binding sites.
  • T²F-Pharm · Pharmacophore models are an accurate and minimal tridimensional abstraction of intermolecular interactions between chemical structures, usually derived from a group of molecules or from a ligand-target complex. Only a limited amount of solutions exists to model comprehensive pharmacophores using the information of a particular target structure without knowledge of any binding ligand. In the T²F-Pharm project, we developed an automated and customable tool for truly target-focused pharmacophore modeling.
  • TeachOpenCADD · TeachOpenCADD is a teaching platform offering tutorials on central topics in cheminformatics and structural bioinformatics. The tutorials contain theoretical background and practical implementations using open source data and software. Implementations are available in two formats: Python-based Jupyter notebooks and GUI-based KNIME workflows.