From discovery to clinic: deep visual proteomics for advanced precision medicine

The individual architecture of tissue holds information that is invaluable for clinical diagnosis and treatment. Molecular approaches such as genomics consult the pathological assessment of patient specimens, but often fall short owing to nonlinear genomic-phenotypic relationships. Immuno-based imaging of tissue sections provides spatial resolution of a phenotype by targeting specific proteins but lacks information depth and requires a-priori knowledge on the analysis outcome. Correlating the visual and proteomic phenotype of single cells while maintaining their localization in the native tissue landscape is hence key to improving precision medicine.

We aim to explore disease-causing proteins and to identify druggable targets in individual patients. To this end, we employed a robust and scalable adaptation of the Deep Visual Proteomics (DVP) technology. DVP integrates high-content imaging, AI-based image analysis, single-cell isolation via high-precision laser microdissection and ultra-sensitive mass spectrometry-based proteomics. By employing cutting-edge artificial intelligence models, our platform enables accurate segmentation and classification of diverse cell types within tissue samples. This advanced phenotyping approach facilitates the automated identification of morphologically distinct regions of interest within each tissue, laying the foundation for targeted analysis and discovery of novel therapeutic targets. Integrating laser microdissection (LMD7, Leica Microsystems) for precise single-cell isolation from minute tissue samples with advanced mass spectrometry (Astral, ThermoFisher) and standardized liquid chromatography (Evosep), our platform achieves a throughput of 80 samples per day, yielding in-depth proteomes of over 6,000 proteins from single-cell isolated phenotypes across diverse diseases and tissues. The generation of proteomic maps based on the original tissue specimens resolves distinct cell types or cell states within AI-annotated regions that represent the phenotypic heterogeneity of a given disease. We demonstrated the versatility and effectiveness of our platform by applying our spatial approach to complex oncology cases across various indications.

In summary, our approach enables robust and scalable spatial proteomics to identify cell-type specific, druggable protein targets across diverse indications. This resolves the cellular heterogeneity of a disease in synergy with clinical pathology and presents clinicians with disease-relevant therapeutic options at the individual level. The aim of precision medicine is to match the right drug to the right patient. Using our spatial proteomic approach, we translate recent technological advances from discovery to the clinic, identifying novel and druggable protein targets.