Explainable AI for inflammation

Artificial Intelligence possesses unprecedented ability in studying and identifying patterns in complex datasets with high dimensions. They effectively overcome the curse-of-dimensionality, a significant obstacle for scientists when analysing several disease features simultaneously. The XAI4Inf group focuses on utilising the variational auto-encoder, a type of generative deep neural network, to transform complex single-cell transcriptomics data into a lower-dimensional latent space. We then employ explainable artificial intelligence techniques to uncover disease sub-populations and gene-gene interactions. Currently, we are conducting training on an advanced architecture that encodes the single-cell transcriptomics of paediatric patients who have rare rheumatological disorders being investigated by the Systems Rheumatology Department of the DRFZ.