Advances in Deep Visual Proteomics With Dr. Andreas Mund
DVP connects microscopy with mass spectrometry to map diseases at the molecular level in single cells.
Deep Visual Proteomics (DVP) integrates high-resolution microscopy, AI-driven image analysis and ultra-sensitive mass spectrometry (MS) to chart protein expression at single-cell resolution while preserving spatial context. The technique is already delivering tangible clinical impact – recently, DVP was used to identify a druggable pathway in toxic epidermal necrolysis, enabling targeted therapy to be administered and patient recovery within weeks.
Dr. Andreas Mund, who leads spatial proteomics research in the laboratory of Professor Matthias Mann at the University of Copenhagen, is one of the researchers that contributed to the development of DVP. In a recent interview with Technology Networks, Mund discussed the development of DVP, key technological milestone and how his team is working to make spatial proteomics more accessible through OmicVision. He also reflects on how DVP is being adopted by academic labs and industry partners worldwide, and where he sees the field heading next – from proteoform-resolved analyses to AI-powered predictive models for precision oncology.
Can you explain how DVP was developed?
Andreas Mund, PhD (AM):
My background lies in high-resolution single-cell imaging. At the Novo Nordisk Center for Protein Research at the University of Copenhagen, I worked extensively with quantitative image-based cytometry (QIBC) – a powerful platform for large-scale cellular phenotyping. While QIBC excelled at identifying distinct cellular behaviors, a critical gap remained: we could observe striking phenotypic differences but lacked insight into the molecular mechanisms driving them.
The core question guiding our work was simple yet profound: why do certain cells behave differently? What molecular processes underlie these variations? How can we connect what we see under the microscope to the functional biology within cells?
Initially, we used MS-based proteomics for discovery, followed by microscopy validation. But I proposed flipping this workflow: start with imaging to pinpoint cells of interest, then apply deep proteomic profiling to decode their molecular identity. Importantly, the collaborative and innovative environment in Mann’s group made this integration possible – and his mentorship and dedicated support were essential throughout our journey.
Developing DVP required assembling a multidisciplinary dream team. We partnered with Professor Peter Horvath in Hungary, who brought cutting-edge AI-driven image analysis. Leica Microsystems co-engineered an automated laser microdissection system for precise single-cell isolation. Dr. Fabian Coscia, then a PhD student in Matthias’ group at the Max Planck Institute (MPI), contributed essential expertise in clinical tissue proteomics.
Simultaneously, our team at MPI co-developed – with Bruker and Evosep Biosystems – the first mass spectrometer capable of true single-cell sensitivity, now commercialized as the timsTOF SCP. In late 2020, we demonstrated unbiased single-cell proteomics, quantifying over 1,200 proteins per cell on a prototype instrument. We stationed one of these prototypes at the Novo Nordisk Foundation Center for Protein Research, enabling the ultra-sensitive measurements that power DVP at single-cell resolution. We then unified these capabilities into a single, seamless platform.
DVP emerged from a simple but persistent challenge in biomedicine: how to connect what pathologists see under the microscope with the molecular machinery driving disease – without losing the spatial context that gives tissue its biological meaning.
MC:
What research need has DVP helped to fulfil?
AM:
Traditional pathology relies heavily on morphological assessment, which, while invaluable, provides limited molecular insight. Conversely, standard proteomic approaches typically require tissue homogenization, which destroys the very spatial architecture that defines cellular neighborhoods and functional interactions.
This loss obscures critical biological information: cell–cell interactions, tissue architecture and the precise cellular neighborhoods that drive disease progression. In cancer, for instance, we can’t distinguish tumor cells from immune or stromal cells within the same sample – let alone identify which populations drive resistance or response.
Spatial context reveals how diseases spread, how immune cells infiltrate tumors and how microenvironments shape clinical outcomes. Without it, we miss therapeutic targets and biomarkers that exist only in specific locations. DVP bridges this divide by delivering deep proteomic data while preserving spatial integrity.
Importantly, DVP is designed as an extension – not a replacement – of existing pathological workflows. Its modular architecture integrates smoothly into clinical practice and evolves with advances across imaging, proteomics and AI.
MC:
What iterations/ updates to the DVP method have been made since it was first developed? Which of these do you think are the most impactful?
AM:
We’ve achieved five major advances:
- Single-cell DVP (scDVP) now quantifies over 4,000 proteins per cell – a 100-fold improvement in sensitivity since our initial proof-of-concept.
- We pioneered multimodal spatial analysis, combining DVP with spatial transcriptomics. In our ovarian cancer study, Dr. Lisa Schweizer used this approach to simultaneously profile thousands of proteins and transcripts across tumor progression stages, revealing unprecedented detail about the tumor microenvironment.
- We integrated multiplexed immunofluorescence with deep proteomics, directly informing immunotherapy development by mapping functional immune–tumor interactions.
- Matthias’ lab extended DVP to proteoform-level analysis and developed nanoPhos, an ultra-sensitive phosphoproteomics method that dramatically boosts detection of phosphorylation sites from minimal samples. By combining lossless SPEC with Fe(III)-NTA enrichment, nanoPhos identifies thousands of phosphosites from tiny inputs – bringing post-translational modification (PTM) mapping into spatial and cell-type resolution.
Our AI capabilities have also matured. We now support supervised learning, where pathologists train algorithms to recognize rare phenotypes like cancer stem cells, and unsupervised learning, which uncovers subtle patterns invisible to the human eye.
Through OmicVision, we’re scaling DVP using computational foundation models. We’ve built a novel H&E-to-proteome pipeline that decodes protein expression directly from standard histology slides via tile-level mass spectrometry. This creates a one-to-one map between histological features and proteomic profiles – enabling AI models to predict molecular states from routine pathology images alone. This breakthrough enables routine pathology labs to access deep molecular insights – transforming standard histology into a gateway for spatial proteomics.
MC:
How are you using DVP in your own research?
AM:
Our most transformative clinical application to date is in toxic epidermal necrolysis (TEN) – a rare, life-threatening condition with 20–25% mortality, often triggered by common drugs like aspirin. Patients suffer widespread skin necrosis, and current treatment (high-dose corticosteroids) causes severe side effects, including vision and hearing loss.
Using DVP, Dr. Thierry Nordmann from our lab discovered hyperactivated JAK/STAT signaling, with elevated STAT1 specifically in affected skin. Recognizing this as a druggable pathway – JAK inhibitors are already used in other inflammatory diseases – we validated the target in mouse models, showing both prevention and reversal of disease.
In a rapid clinical collaboration in China, seven TEN patients received JAK inhibitors. Results were dramatic: disease progression halted within two to four days, skin healing began within a week and all patients recovered fully in two to three weeks.
We completed the full discovery-to-clinic cycle in just 18 months – a landmark for spatial omics and a new paradigm for translational medicine. This work, published in Nature, marks the first successful clinical application of spatial proteomics.
Our collaboration with Dr. Ernst Lengyel at the University of Chicago showcases DVP’s power in ovarian cancer. Lengyel – a leading gynecologic oncologist and long term collaborator – partnered with us to conduct the first spatially resolved molecular mapping of borderline ovarian tumors as they progress to invasive carcinoma.
Using DVP combined with spatial transcriptomics, Schweizer (now at OmicVision) led profiling across tumor stages – from benign serous borderline tumors to invasive, chemotherapy-resistant low-grade serous carcinoma. We identified NOVA2 as a protein that is uniquely expressed in invasive cancers, potentially driving malignant transformation.
The study revealed promising drug targets and demonstrated that dual therapy with milciclib and mirvetuximab significantly reduced tumor growth in mice. As Lengyel noted, this work exemplifies how cross-institutional collaboration can accelerate therapy development for challenging diseases – and lays the foundation for upcoming clinical trials.
MC:
What is your favourite example of how another lab is using DVP?
AM:
DVP adoption is accelerating as labs recognize its translational potential. Its modular design allows adaptation to diverse questions – from cancer to neurobiology. Matthias’ lab launched training programs together with the labs of Professor Peter Horvath and Dr. Fabian Coscia, including the first DVP workshop in Vienna, to help labs implement the workflow.
While technically demanding, DVP’s greatest strength lies in fostering collaboration: pathologists identify regions of interest, and researchers decode their molecular logic – creating insights no single discipline could achieve alone.
Fabian Coscia’s lab offers a great example of DVP’s versatility. His team combined electrophysiology with DVP to investigate pain-sensing neurons*, asking: which proteins define neuronal diversity in pain pathways? By isolating electrophysiologically distinct sensory neurons and profiling their deep proteomes, they uncovered novel proteomic markers and subset-specific signatures – opening new avenues for targeted pain therapeutics.
Beyond academia, pharmaceutical companies and cancer centers are actively exploring DVP for drug discovery. Its ability to deliver hypothesis-free, spatially resolved proteomics makes it uniquely powerful for precision medicine.
Ultimately, DVP isn’t just a method – it’s a new analytical paradigm, reshaping how we study tissue biology and disease mechanisms.
*This article is based on research findings that are yet to be peer-reviewed. Results are therefore regarded as preliminary and should be interpreted as such. Find out about the role of the peer review process in research here. For further information, please contact the cited source.
MC:
How do you think DVP could continue to evolve?
AM:
OmicVision Biosciences is driving DVP toward accessibility and scalability. While academic labs advance the science and train new users – like our joint DVP workshop with the Coscia and Horvath labs at the ESCP conference in Vienna – clinical impact demands different infrastructure.
OmicVision partners directly with pharma, hospitals and cancer centers to translate innovations into patient benefit. We’re eliminating the need for labs to build or maintain the full DVP pipeline in-house, making advanced spatial proteomics accessible through collaboration – regardless of local resources – and enabling widespread implementation within the coming years.
AI and machine learning will be transformative. As we generate high-resolution protein maps of diseased tissues, these data become molecular overlays on traditional pathology, guiding diagnosis and treatment.
The next frontier is proteoform-resolved DVP. With nanoPhos, we can now detect thousands of phosphorylation sites from minute samples – extending spatial proteomics into the dynamic world of PTMs.
Another major evolution is integrating complementary omics: untargeted discovery with targeted validation. This hybrid approach delivers both breadth and depth – revealing global proteomic landscapes while zooming in on spatially precise mechanisms.
We’re also building predictive models that link spatial protein patterns to clinical outcomes. Within a few years, these models will enable precision oncology applications across multiple diseases.
Critically, the field is on the cusp of a throughput revolution. Advances in short-gradient MS and multiplexed sample preparation are poised to increase proteomics capacity by orders of magnitude – enabling the measurement of hundreds to thousands of samples per day. When integrated with DVP, this leap in scale will make spatially resolved proteomics feasible not just for discovery, but for routine clinical screening and longitudinal monitoring.
DVP will eventually become standard in precision medicine – especially in oncology and neurology. By revealing the spatial logic of cellular function, it will redefine diagnostics and therapy. We expect early clinical adopters within the next 5 years and broad implementation within 5–10.
