Professor Krawitz is a distinguished expert in the field of medical genetics and bioinformatics. He received board certification in Medical Genetics in 2016 and completed a postdoctoral lecture qualification in Human Genetics the same year. From 2009 to 2016, he completed a residency at Charité, Berlin. Prior to this, he attended medical school at the Technical University Munich from 2001 to 2008 and studied Physics at Ludwig Maximilian’s University from 2001 to 2007.
Professor Krawitz’s scientific and professional career includes significant roles such as:
2020- Faculty of Immunosensation Cluster of Excellence
2018- Faculty of Bonn-Aachen International Center for Information Technology (B-IT)
2017- Director of the Institute for Genomic Statistics and Bioinformatics, University Bonn,
2016-2017: Group Leader Clinical Bioinformatics, Charité Medical School, Berlin
2015-2022: Chief Science Officer, FDNA
2013-2017: Principal Investigator, GPI-anchor deficiencies
2009-2013: Postdoctoral research fellow in the lab for Computational Biology of
Prof. Dr. Peter Robinson at Charité Medical School, Berlin
2011- Cofounder of GeneTalk
2006-2007: Research fellow in the laboratory for Computational Biology of
Prof. Dr. Ilya Shmulevich at Institute for Systemsbiology, Seattle
2003-2006: Research fellow in the laboratory for Neurodegenerative Disease of
Prof. Dr. Christian Haass at University Hospital Munich
In addition to research, Professor Krawitz has been actively involved in teaching:
Since 2017 Life Science Informatics, Bonn-Aachen, International Center for Information
Technology, B-IT
2016-2017 Medical Genomics for Students of Bioinformatics, Free University Berlin
2009-2015 Human Genetics for Medical Students at Charité, Berlin
His fields of interest include the analysis of whole genome data with a special focus on mutation prioritization and next-generation phenotyping, particularly the application of artificial intelligence in precision medicine.
GestaltMatcher: medical image analysis with AI in rare diseases
Next-generation phenotyping approaches have demonstrated high accuracy in suggesting diagnoses based on dysmorphic facial features. We propose a novel approach using convolutional neural networks (CNNs) to simultaneously infer Human Phenotype Ontology (HPO) terms and classify disorders based on syndromic faces. By integrating HPO extraction and disorder prediction, our method enables cross-validation of results, improving diagnostic robustness. We aim to enhance both diagnostic support and explainability in AI-driven rare disease classification.
Through GestaltMatcher Database, we set up a large-scale effort to annotate the presence and absence of HPO-terms. As annotation of HPO-terms is a subjective process, we allowed multiple experts to annotate the same images to improve the overall robustness. By simultaneously learning to predict disorder and infer HPO-terms, we can cross-reference inferred terms with disorders to (in)validate predictions. Further, since HPO-terms inherently carry spatial information, we can gain new insights into the predicted disorders.
At the time of writing more than 1000 images, across 10 disorders were annotated with 72 common HPO-terms for the disorders, alongside any term the annotators considered important. We trained several CNNs and compared the performance when predicting HPOs, disorders, and a combination of the two. The latter allowed us to combine the predictions to further improve performance and explainability.
Our proposed approach allows for more and better analyses to be conducted straightforwardly. Further, it inherently improves the explainability of our AI models, improving their usefulness in clinical settings. Additionally, it reduces the subjectiveness of labeling HPO-terms manually and allows clinicians to work more efficiently.