SPP 2363 on “Utilization and Development of Machine Learning for Molecular Applications – Molecular Machine Learning”

From Fundamentals to Application and Beyond
“Today the computer is just as important as tool for chemists as the test tube. Simulations are so realistic that they predict the outcome of traditional experiments.”
The Royal Swedish Academy of Science, 2013

Staying abroad with the SPP2363 – Machine Learning at Vanderbilt University, USA

A short report by Fabian Liessmann, Group of Prof. Jens Meiler, University of Leipzig


© Fabian Liessmann, Uni Leipzig
  • © Fabian Liessmann, Uni Leipzig
  • © Fabian Liessmann, Uni Leipzig

Thanks to the scholarship for PhD students to stay abroad of the SPP2363, I had the opportunity to expand my research projects for several weeks at the Vanderbilt University, Nashville, Tennessee, USA. During that time, I worked together with several members of our partner lab, and especially, with the research assistant professor Benjamin (Ben) Brown, who specialized in the development and application of computer-aided drug design methods. Ben Brown is one of the main developers and contributors of the BioChemical Library, in short BCL, which employs machine learning algorithms for more than 15 years. Furthermore, he is an expert for optimizing drug candidates with both traditional and deep learning methods. In many fruitful discussions, we evaluated the current paradigm shift in the field and discussed the future of machine learning, especially for small molecules. Furthermore, during a special session about “Recent Developments and Technologies in ML”, I had the opportunity to listen to presentations from various researchers and directly discuss it with the lab members. A highlight was definitely the talk “Towards Data-Centric Graph Learning for Real-World Applications”  from Assistant Professor Tyler Derr from Vanderbilt, where he presented his current research projects and machine-learning approaches.

This internship in Nashville gave me valuable insights outside of my day-to-day research environment and work. Of course, the personal and direct communication facilitates the exchange of research ideas, especially during hiking or group events in Nashville. The international lab, its environment and the welcoming/open nature of America captured and embraced me and round off an exciting time. I am really grateful to the SPP2363 for this unique and rewarding opportunity.

Annual meeting of the SPP 2363 in Jena

© Dr. Stefan Zechel, FSU Jena

This year’s annual meeting of the SPP 2363 took place at the Friedrich-Schiller University of Jena from the 12th to the 14th of September 2023. Several lectures and a poster session updated everyone on the current status of the research projects as well as providing an interactive setting for exchanging and discussing new ideas. In addition to the research project updates from all members of the SPP 2363, there were also workshops on gender & diversity in science as well as a hands-on introduction to robot-based chemistry. Furthermore, two speakers from industry gave fascinating insights into automation and data management. Overall, the meeting was a great success by strengthening connections, exchanging ideas and forming new synergies.

© Leon Schlosser, Uni MS

The PhD students of the SPP 2363 participated in a workshop on scientific presentations on the 22nd and 23rd of August 2023 in Münster. Led by the experienced trainer Dr. Alexander Britz the students enhanced their skills for effective scientific communication with a special focus on oral and poster presentations. In interactive and fun exercises the participants learned techniques to present their scientific results in a structured and appealing way.

Combining machine learning and physical knowledge: Junior research group develops models for industrial production

Junior Professor Dr. Fabian Jirasek conducts research on machine learning in process engineering.
© RPTU/Koziel

Fabian Jirasek is granted the new Emmy Noether Junior Research Group "Hybrid Thermodynamic Models" by the German Research Foundation (DFG), which will be established at RPTU Kaiserslautern. The group will develop novel thermodynamic models to predict the fluid properties of mixtures. The knowledge of these properties is essential for the chemical industries, e.g., for process optimization and establishing new production routes that rely on renewable instead of fossil raw materials.
Mixtures are omnipresent when producing chemicals and medicines, but also in developing new batteries, e.g., for e-cars. "Understanding the properties of these mixtures is of central importance for basically all processes in chemical engineering, from the reaction to the purification of the target products," says Fabian Jirasek, who is a Junior Professor for Machine Learning in Process Engineering at RPTU Kaiserslautern. "They are the basis for design and optimization of efficient processes."
Investigating all conceivable combinations of substances and the influence of parameters, such as temperature and pressure, in laboratory experiments is not feasible due to the abundance of possibilities. "Therefore, we rely on our models to also predict the fluid properties of non-measured substances and mixtures as well as at non-measured states," explains Jirasek. In thermodynamics, such models have been established for decades. But thanks to machine learning (ML), a subfield of artificial intelligence, research today has entirely new possibilities. "ML techniques will revolutionize thermodynamic modeling," Jirasek is certain. In the new project, Jirasek and his team will develop hybrid models combining machine learning and physical modeling. "We assume that machine-learning methods will require significantly less data by exploiting the available physical knowledge," he continues. "This will also create trust and acceptance for the novel models if they or their predictions satisfy known physical laws."
The DFG's Emmy Noether Programme is aimed at outstandingly qualified young researchers. The goal is that they can thereby qualify for a professorship by independently leading a junior research group.
Fabian Jirasek studied bioengineering at the Karlsruhe Institute of Technology and earned his doctorate in thermodynamics at Kaiserslautern. He then researched at the University of California at Irvine before moving to TU Munich. Since the fall of 2020, he has held the Junior Professorship for Machine Learning in Process Engineering at RPTU Kaiserslautern, funded by the Carl Zeiss Foundation.

Review: Navigating large chemical spaces in early-phase drug discovery

© ScienceDirect®, Elsevier B.V.

The use of chemical spaces for drug discovery has become a major focus in the scientific community, highlighting their growing importance. In a recent review paper, co-authored by Malte Korn from the SPP 2363 and published in Current Opinion in Structural Biology, a detailed examination of the current state of chemical spaces is given. The article covers the fundamentals of large chemical spaces and their importance for drug discovery making it an excellent introductory source for non-experts.
So, if you are new to this field start your journey through chemical spaces with this open-access review today!

Cover Grimme 2023
© AIP Publishing, J. Chem. Phys.

Cover on The Journal of Chemical Physics

The latest semiempirical tight-binding method for electronic structure calculations by the Grimme group (S. Grimme, M. Müller, and A. Hansen) has made it to the cover of The Journal of Chemical Physics! The method, named PTB, is excellently suited for the generation of descriptors in the context of modern machine-learning approaches based on quantum-chemical features. As PTB is available for all elements up to Radon (except for lanthanides) and is a non-self-consistent two-step method, it provides consistent and robust results throughout large parts of the chemical space. The method will be publicly available in the open-source xtb program package - give it a try!

5th International Mini-Symposium Molecular Machine Learning, January 19th 2023

Yesterday, the fifth edition of the international symposium series on "Molecular Machine Learning" took place virtually. Over 200 participants joined the symposium and made it yet again a very special event. The invited speakers Núria López (ICIQ), Kim Jelfs (Imperial College London), Tim Cernak (University of Michigan) and Sarah Reisman (Caltech) gave fascinating insights into their research and highlighted how data-driven approaches can be utilized for computer-aided synthesis planning, molecular design, automation and drug discovery.

5th International Mini-Symposium Molecular Machine Learning, January 19th 2023

© Felix Katzenburg, Uni MS


The fifth edition of the international symposium series on "Molecular Machine Learning" organized by Prof. Frank Glorius will take place on January 19th, 2023 from 3:00 pm. The virtual conference series brings together leading scientists from fields including computer-aided synthesis planning, data-driven molecular design, automation and AI-enabled drug discovery. Speakers at the symposium will be: Núria López (ICIQ), Kim Jelfs (Imperial College London), Tim Cernak (University of Michigan) and Sarah Reisman (Caltech).

SPP 2363 Kick-off Symposium

Kick-off Meeting

The SPP 2363 Kick-off Symposium took place at the German National Academy of Sciences Leopoldina in Halle from October 17th to October 21st. With PhD talks and posters on the various applications of molecular machine learning anchored in the priority program and the opportunity to build an even stronger the community, the event was a great success. Talks from industry and researcher provided a great setting for exchanging ideas and setting goals for the future development of the program. We were also very pleased that Tiago Rodrigues (University of Lisbon) accepted our invitation and joined the symposium as the first SPP 2363 Mercator Fellow. The event was supported by the German National Academy of Sciences Leopoldina and the DFG.

SPP 2363 officially started

The DFG has selected suitable projects and the official funding letters were received. Let' get started, everyone!
Interested students, kindly contact the respective principal investigators.

© 2022 American Chemical Society

In C&EN's June cover story, Matthias Rarey highlights how computational tools help to navigate chemical spaces and virtual libraries in the search for new drugs.

Connecting chemical building blocks allows drug hunters to explore a much bigger chemical space than before. The challenge is to narrow this field of compounds to something manageable. To do that, chemists are turning to new computational tools to navigate this increasingly huge chemical universe, and they are combining technologies. Experts say these new approaches should speed up the identification process, and industry is investing time and money to optimize the hunt.

“We must learn to understand chemistry as data science”

Frank Glorius and Philipp Pflüger talk about the new field of research “Molecular Machine Learning".

“Molecular Machine Learning” (MML) is a new branch of research with the potential to change chemical research. Prof. Frank Glorius, coordinator of the new Priority Programme “Molecular Machine Learning” (SPP 2363), funded by the German Research Foundation (DFG), and Philipp Pflüger, who is working on his PhD in Chemistry and helped to develop the programme, explain in this interview with Christina Hoppenbrock what MML means, what opportunities and challenges this new field of research presents, and what working in chemistry will be like in tomorrow’s world.

Please follow us on Twitter

© Philipp Pflüger/Uni MS; Twitter, Inc.