Graphs further underlie many complex biological processes such as protein interactions, which are important to understand in detail, for example for drug development and repurposing, but which are difficult to assess empirically with a limited number of parameters. Researchers have therefore resorted to creating artificial graphs which resemble what has actually been observed—to better understand the underlying distribution of such complex data and deduce what is not apparent in the data.
Bastian Rieck, Principal Investigator at AI for Health and at the Helmholtz Pioneer Campus in Munich, his PhD student Jeremy Don Wayland, and their collaborators have now leveraged Bastian’s expertise in so-called topological data analysis to develop a new method that can evaluate approaches to create such artificial graphs. This work has now been highlighted as groundbreaking research by DeepAI, which regularly identifies the most popular research in artificial intelligence around the world: The evaluation of graph-creating approaches is important since the underlying distribution of the newly created graphs is not known; it is, however, essential that the artificial graphs reflect the distribution of the real-world data so that the newly generated data can be used to simulate complex systems. For example, researchers could use this method to better understand the spread of infectious disease, such as during the on-going COVID-19 pandemic: While one pandemic delivers limited information about pathogenic spread, the generation of graphs with the same underlying distribution can help us simulate likely scenarios of future outbreaks and therefore significantly contribute to our pandemic preparedness. Bastian's research is exploring new avenues to make such approaches more reliable, for the ultimate benefit of our society.
Biomedical AI: topology and curvature accelerate biomedical data analytics
