Learning the Rules of Molecular Design

Mentored by Peter F Stadler, Carsten Lutz
at Leipzig University

Design problems in chemistry can be phrased as the inverse of a structure and or property prediction problem in the following sense: Given a function f that returns a vector of properties for each object x, and a target property vector g, we are interested in finding objects x such that |f(x)-g| < t for some tolerance t. This is often phrased as an optimization problem to minimize |f(x)-g|. Instead, we are interested here in learning rules for composing x from constituent fragments. As a concrete application domain, we will focus on the design of molecules, represented as graphs labeled by atom and bond types. The focus of the project is on learning hierarchical procedures that integrate with a generative approach for constructing sub-solutions. This will involve the combination of methods from theoretical computer science with recent developments in generative AI and a strong emphasis on explainability due the focus on design rules rather than just the designed objects themselves.

Work Environment

The primary place of work will be the bioinformatics groups at Leipzig University, which provides an interdisciplinary work environment with faculty, staff, and peers from different backgrounds in the sciences, from Computer Science and Mathematics to Chemistry, Physics, and Biology. The project will be conducted in close cooperation with the Foundations of Knowledge Representation group, providing in-depth mentoring in Theoretical Computer Science. Of course, the usual resources of office space, administrative support, and access to adequate computing facilities will be available.

SECAI offers a first-class environment for advancing your career. You can work with internationally renowned researchers and benefit from the school’s strong networks in industry and research. The graduation of highly qualified researchers is a central project goal in SECAI and doctoral students receive strong support for their professional and personal development.

Prerequisites

• Very good university degree (M.Sc. or an equivalent) in Computer Science, Mathematics or Chemistry with a strong interest and at least a basic education in the other fields
• Keen interest in working in an interdisciplinary work environment on a topic that crosses traditional boundaries between research fields
• Interest in advancing methods and their underlying formal framework