Latest updates

Our Christmas party this year was a success! With a lot of homemade food, a decorated Christmas tree and a secret santa, we celebrated the end of the year together. We have grown a lot this year and it was wonderful to get to know everyone better.

We look forward to another great year!

Read more

Lisa completed her bachelor and master studies in Bioinformatics at Saarland University. After being a research assistant at our lab for almost a year, she now joined our team as PhD candidate. Her resarch focuses on the development of trustworthy ML models. Particularly, as part of the EU-funded RADAR project, she develops interpretable and reliable ML methods that are used to create sustainable plastic alternatives.

Read more

We attended this year’s German Conference on Cheminformatics in November held in Bad Soden am Taunus. We were happy to present our work on multiple posters:

• Floriane Odje: Morphological data analysis: From descriptor development to predictive modelling
• Afnan Sultan: Domain adaptation as a computationally efficient approach for improving transformer models for molecular property prediction
• Yanyuan Zhang: Read-Across the Targetome – An integrated structure- and ligand-based workbench for computational design of novel tool compounds
• Hamza Ibrahim: MolDockLab: Data-driven workflow for best balanced consensus docking pipeline for hit identification
• Paula Kramer: Active learning for fragment-based kinase inhibitor design using docking
• Michael Backenköhler: Structural activity prediction models recover known kinase binding modes
• Katharina Buchthal: Novel kinase ligand generation using subpocket-based docking
• Erika Primavera: Why is Miransertib effective against the AKT1-E17K mutation?

Antoine Lacour was accepted for a research talk where he presented his research on “DockM8: All-in-one open-source platform for consensus virtual drug screening”. We are happy to announce that Paula won the GCC poster award and Michael received an honorable mention for his poster.

Read more

Our group is growing so fast, we had our own group retreat to spend some time together. We started with group games at the office and ended the day with pizza and drinks in the city. We all had a lot of fun and got to know each other much better.

To be repeated soon!

Read more

This year’s RDKit user group meeting was organized by Sereiner Riniker’s lab in Zurich, Switzerland on September 11th-13th. We used this opportunity to present our current projects during the poster sessions:

• Floriane Odje: Morphological data analysis: From descriptor development to predictive modeling
• Paula Kramer: Active learning for fragment-based kinase inhibitor design using docking
• Hamza Ibrahim: MolDockLab: Data-driven workflow to find best balanced consensus docking pipeline for hit identification

Additionally, Afnan Sultan gave a talk about the current state of transformers for property prediction and Antoine Lacour introduced us to his consensus docking tool called DockM8 in his lightning talk.

Read more

We are happy to announce the publication of our review article “Transformers for Molecular Property Prediction: Lessons Learned from the Past Five Years” in the Journal of Chemical Information and Modeling: https://doi.org/10.1021/acs.jcim.4c00747.

In this review, we take a closer look at how transformer models have been used for molecular property prediction over the past five years. We examine the different models and try to answer some of the key questions that come up when developing these models: How much pretraining data is needed? Which model architectures and pretraining objectives are promising? And how can we meaningfully compare the performance of different models? We also highlight gaps in the current research and the need for more standardized data splitting and robust statistical analysis when benchmarking molecular transformer models.

Read more

As part of the ICML conference in Vienna, we attended the Machine Learning for Life and Material Science workshop in Vienna, Austria on July 26th. Michael and Joschka presented their poster on the current updates on their kinodata-3d project by investigating the explainability of their structural model for binding affinity prediction. Paula presented an approach for fragment-based kinase inhibitor design using active learning on her poster. We especially enjoyed meeting fellow researchers working in the intersection of machine learning and drug discovery.

Read more

We recently published a literature review in Frontiers in Toxicology titled “Unleashing the Potential of Cell Painting Assays for Compound Activities and Hazards Prediction.”
This paper is part of the research topic “Leveraging Artificial Intelligence and Open Science for Toxicological Risk Assessment.”
In this review, we highlight how single-cell-level data from cell painting assays can be combined with structural information to predict compound activities for various human-relevant disease endpoints and to identify underlying modes of action using machine learning and deep learning. We also discuss the current challenges associated with this approach.
Link to the paper: https://www.frontiersin.org/journals/toxicology/articles/10.3389/ftox.2024.1401036/full

Read more

Hamza has successfully defended his master’s thesis titled “MolDockLab: A Data-Driven Approach to Optimizing Docking Tools, Scoring Functions, and Ranking Methods for Targeted Applications.”

In his research, he developed MolDockLab, a data-driven approach for optimizing the combination of docking tools, scoring functions, and ranking methods for specific targets. Using energy-coupling factor (ECF) transporters as a case study, his approach identified 18 compounds from the in-house HIPS library for further in vitro assays.

Read more

We are happy to announce that our paper “Reliable anti-cancer drug senstivity prediction and prioritization” was finally published in Scientific Reports. Tackling the challenge of reliable ML predictions in medical applications, we developed a novel approach for predicting and prioritizing anti-cancer drug responses with guaranteed certainty levels, a unique contribution to the field. This research was conducted in collaboration with Kerstin Lenhof, Lea Eckhart, Lisa-Marie Rolli, and Hans-Peter Lenhof from the Lenhof chair.
For more information check out our LinkedIn post or directly read our paper.

Read more