Helmholtz AI project call showcase: AI-based development of precision medicine for Alzheimer’s disease

Using deep learning approaches that combine proteomic and clinical data with 3D spatial proteomics amyloid plaque data for generating diagnostic tools allowing Alzheimer's disease precision medicine.

How can deep learning help in the research of Alzheimer's disease? Find out in today's Helmholtz AI project showcase, a collaboration between scientists from Helmholtz Munich and the German Center for Neurodegenerative Diseases (DZNE).

Could you introduce yourself, giving your affiliation, area of work, and of course, the project title?

Dr. Ali Ertürk, Director of iTERM (Institute for Tissue Engineering and Regenerative Medicine) at Helmholtz Munich, focuses on developing and implementing technologies to enable personalised treatment of complicated diseases. Together with Prof. Dr. Stefan Lichtenthaler from the German Center for Neurodegenerative Diseases (DZNE), he is working on the project ‘DEEPROAD: AI-based development of next generation diagnostics and precision medicine for Alzheimer’s disease’.

In simple words, what specifically is your project about? And, how and why do you think it is a high risk, high gain endeavour?

Alzheimer’s disease (AD) affects 50 million people worldwide, but causal treatments are not yet available. Recent phase 3 clinical trials with amyloid-targeting drugs were unsuccessful. It is now clear that successful trials require a stratification of AD patients. To generate the diagnostic tools allowing AD precision medicine, we propose to use deep learning approaches combining proteomic and clinical data with 3D spatial proteomics amyloid plaque data that we believe can reveal hidden patterns of AD within the different subgroups that can help in future successful treatments. Both the clinical parameters and proteomics data that we are using to stratify AD patients are heterogeneous, which might encumber both the discriminative power of our clustering approach and so the generalisability of its results. In this project we are also combining spatial information to the proteomics data which is a novel approach [Bhatia et al. 2021, biorxiv]. Using the spatial dimension we can be able to study the inter-cellular correlations among Aβ plaques between different AD subgroups. Studying within tissue interactions is quite challenging especially with the high dimensionality of the proteomics data, however this can enrich the classification of AD pathology based on the Aβ plaques with precise proteome and location in the brain, in an unbiased way. In the end we want to share our data and models in an accessible platform such that the different research centres can make use of transfer learning by fine-tuning the models to their datasets.

How important has the Helmholtz AI funding and platform been to carry out this project?

The Helmholtz AI funding helps in supporting and providing the resources that can make this project possible by providing the necessary funds including staff expenses, computational needs including storage and processing power, as well as equipment and material expenses for additional proteomics measurements.

Figure: DEEPROAD project outline.