From Protein Structure Prediction to Explainable AI in Oncology

Mentored by Jakob Nikolas Kather & Jens Meiler
at Else Kroener Fresenius Center for Digital Health or Leipzig University, Institute of Drug Discovery

Artificial intelligence (AI) is rapidly transforming not only basic research, but also clinical medicine. AI can process complex data such as gene sequences, protein structures and clinical image data. However, a common limitation of AI approaches in clinical research is that the results are not easily interpretable by the patient and often difficult to explain or present. Providing options to transform data generated by AI to non-experts is an unaddressed need in research and daily clinical routine. The recent breakthrough in protein structure prediction provides a unique opportunity to create workflows from data driven AI research to molecular hypotheses and with that to visually attractive representations. In this project, the PIs Jakob Nikolas Kather and Jens Meiler bring their expertise together to address this need. In Jakob Kather’s group deep learning is used to circumvent labor-intensive molecular analysis but retrieve information from readily available histological samples to predict clinical outcome. Jens Meiler focuses on understanding the molecular foundation of protein function through structural modeling using the Rosetta software package and deep learning methods. Structural modeling results in all-atom molecular presentations of proteins of interest. Together with macromolecular data from the Kather lab, a comprehensive model of the disease state will be produced. These will serve as basis to produce visualizations, verbal presentation and explanations that make complex AI analysis explainable to non-experts. In this project, new methods will be developed and their impact on understandability for non-experts will be quantitatively analyzed.

Work environment

Depending on final arrangements, final appointment can be at TU Dresden or LU Leipzig. We expect the trainee to travel frequently between both laboratories. The Interdisciplinary Center for Bioinformatics in Leipzig and cooperate with the machine learning and AI experts at the ScaDS.AI in Leipzig and Dresden. You will have access to machine-learning HPC resources of the ScaDS.AI Center for Scalable Data Analytics and AI Dresden/Leipzig, offering state-of-the-art CPU (Intel and IBM POWER 9) and GPU (Nvidia A100 and V100) resources, as well as high-performance computing and storage systems for computational experiments. At the Else Kroener Fresenius Center (EKFZ) at TU Dresden, you will be part of an interdisciplinary, diverse team of scientists with professional training in biology, medicine, engineering and computer science. At University Hospital Dresden, you will have access to clinical real-world data in a computing environment within the clinical infrastructure.

Prerequisites

The candidate will need a solid background in bioinformatics or structural biology as well as proficiency in scientific programming with Python and interest in data analysis. Prior expertise in machine learning, genomics or transcriptomics, and deep learning will be an advantage.

Further details on the requirements and application process can be found in SECAI's announcement for open PhD positions in 2024.