ASSAF MAGEN, PhD, Entrepreneur & consultant Computational Biology


Entrepreneur & Consultant
Computational Immunology


I am a former computational biology assistant professor in cancer immunology. I am currently developing innovative AI-enabled solutions for high-dimensional omics data analysis and visualization. I also provide strategic and analytical consulting services to industry and academic partners.

My research focused on discovery of immune cell interaction patterns controlling T cell differentiation fate and understanding the impact these patterns have on response or resistance to immunotherapy. Targeting immune checkpoints such as CTLA-4 and PD-1/PD-L1 has revolutionized cancer treatment, leading to vast efforts to block numerous additional checkpoints and explore combination approaches. However, a more systematic approach to discover the key rate limiting steps of effective immune cell activation and differentiation are crucially needed. My recent work leveraged deep spatial mapping of biopsies and surgical resections to discover resistance mechanisms to PD-1 and CTLA-4 blockade, therapeutic targets that may prevent or reverse them, and predictive biomarkers of response.
Read more here.

Assaf Magen, Ph.D.

News & Media

SDxAI Hackathon Finalist
Our Mendel AI project made it as a finalist. Our goal was to empowers biologists with computational skills to visualize and analyze high-dimensional genomics data through a user-friendly, no-code interface. The tool generates accurate analysis and visualization code from natural language prompts, presenting results in the chat interface. It marks a significant step towards AI-assisted data analysis for biotech and pharma R&D scientists. Future plans include fine-tuning the LLM model, enabling more sophisticated execution, and expanding analysis capabilities. Read more about the hackathon work here.

SDxAI Hackathon
I am forming a team focused on AI-enabled computational biology data science and visualization at the SDxAI Hackathon. SDx is the next-gen startup ecosystem for builders in San Diego - is hosting its inaugural event this summer focused on Generative Artificial Intelligence, the SDxAI Hackathon. We're taking over the beautiful UCSD Design & Innovation Building on July 15-16 for this 2-day, in-person event. Join 100+ fellow hackers for workshops, mentor sessions, and plenty of time for hacking! More about the project and vision here, or learn more about my previous projects here.

Study Supports Our Immune Triad Hypothesis
A new preprint provides support to our recent hypothesis that intratumoral immune triads are critical for effective anti-tumor immunity. The study by Gabriel Espinosa-Carrasco and senior author Andrea Schietinger finds that adoptive T cell therapy relies on CD8+ T cells, CD4+ T helpers, and mregDC triads located within tumors. This independent study is exciting and suggests that our proposed mechanism could apply across cancer immunotherapies.

Spotlight on ACIR
Magen and Hamon et al. analyzed surgically resected tumor lesions and matched, non-involved adjacent liver tissues from HCC patients with T cell-rich tumors to identify molecular correlates of response to neoadjuvant ICB therapy. PD-1hi effector CD8+ T cells and CXCL13+ TH cells were preferentially clonally expanded in tumors of responders, whereas a subset of T cell-rich tumors that failed to respond to anti-PD-1 showed increases in PD-1hi terminal CD8+ T cells and Tregs. T cell clones that expanded upon PD-1 blockade were present in tumor lesions before treatment, and PD-1hi progenitor CD8+ T cells were enriched in close proximity to mregDC/CXCL13+ TH niches in responders.

Study underscores broad applicability of Vizgen to advance human health
Vizgen, the life science company dedicated to improving human health by visualizing single-cell spatial genomics information, today announced publication of a study in the June issue of Nature Medicine conducted by Drs. Assaf Magen and Pauline Hamon in the Miriam Merad lab at the Icahn School of Medicine at Mount Sinai, in collaboration with Regeneron Pharmaceuticals, Inc . The spatial characterization described in this study was enabled by the company’s spatial transcriptomics MERSCOPE ® Platform, which simplifies and empowers the entire workflow from sample preparation through data visualization.

Nature Medicine Publication
Despite no apparent defects in T cell priming and recruitment to tumors, a large subset of T cell rich tumors fail to respond to immune checkpoint blockade (ICB). We leveraged a neoadjuvant anti-PD-1 trial in patients with hepatocellular carcinoma (HCC), as well as additional samples collected from patients treated off-label, to explore correlates of response to ICB within T cell-rich tumors. We show that ICB response correlated with the clonal expansion of intratumoral CXCL13+CH25H+IL-21+PD-1+CD4+ T helper cells (“CXCL13+ TH”) and Granzyme K+ PD-1+ effector-like CD8+ T cells, whereas terminally exhausted CD39hiTOXhiPD-1hiCD8+ T cells dominated in nonresponders. CD4+ and CD8+ T cell clones that expanded post-treatment were found in pretreatment biopsies. Notably, PD-1+TCF-1+ (Progenitor-exhausted) CD8+ T cells shared clones mainly with effector-like cells in responders or terminally exhausted cells in nonresponders, suggesting that local CD8+ T cell differentiation occurs upon ICB. We found that these Progenitor CD8+ T cells interact with CXCL13+ TH within cellular triads around dendritic cells enriched in maturation and regulatory molecules, or “mregDC”. These results suggest that discrete intratumoral niches that include mregDC and CXCL13+ TH control the differentiation of tumor-specific Progenitor exhasuted CD8+ T cells following ICB.

Presentation at the Keystone Symposia Learning from the Patient: Reverse Translation
- Padmanee Sharma, University of Texas MD Anderson Cancer Center, From the Clinic to The Lab: Investigating Myeloid Cells and Epigenetic Pathways that Drive Resistance to Immune Checkpoint Therapy
- Christian U. Blank, Netherlands Cancer Institute - NKI-AVL, Biomarkers of Response to Neoadjuvant Checkpoint Blockade
- Assaf Magen, Icahn School of Medicine at Mount Sinai, Intratumoral DC/T Helper Niches Enable Local Reactivation of CD8 T Cells upon PD-1 Blockade
- David R. Glass, Fred Hutchinson Cancer Center, Integration of Multi-omic Data into Immune Modules Predictive of Response to Checkpoint Blockade in Cutaneous T Cell Lymphoma

Invited talk/Faculty visit
Presentation at the James P. Allison Institute at MD Anderson Cancer Center

Presentation at the SITC 37th Annual Meeting: Presidential session
- Metagenomic sequencing reveals unique gut microbial features associated with tertiary lymphoid structures in response to immune checkpoint blockade in solid cancers, Manoj Chelvanambi, PhD – The University of Texas MD Anderson Cancer Center
- Intratumoral mregDC and T helper niches enable local reinvigoration of CD8 T cells following PD-1 blockade, Assaf Magen, PhD – Icahn School of Medicine at Mount Sinai
- Expert Discussant Thomas F. Gajewski, MD, PhD – University of Chicago
- Impact of Tet2-mutant clonal hematopoiesis on solid tumor immunology and response to checkpoint blockade, Shelley Herbrich, PhD – The University of Texas MD Anderson Cancer Center
- Deficiency of metabolic regulator PKM2 activates the pentose phosphate pathway to generate TCF1+ progenitor CD8 T cells to improve efficacy of PD-1 checkpoint blockade, Geoffrey Markowitz, PhD – Weill Cornell Medicine

Presentation at the CRI-ENCI-AACR International Cancer Immunotherapy Conference; Spatial, Temporal and Computational Approaches for Understanding Tumor Immunity
- Haydn T. Kissick, Emory University School of Medicine, CD8 T-cell activation in cancer is comprised of two distinct phases.
- Benjamin G. Vincent, UNC Lineberger Comprehensive Cancer Center, Cancer immunogenomics research with the CRI iAtlas web portal
- Ronald N. Germain, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Gaining Insight into Tumor Immunity By Combining New Highly Multiplex 2D and 3D Imaging with Analytical Tools
- Assaf Magen, Icahn School of Medicine at Mount Sinai, Intratumoral mregDC and CXCL13 T helper niches enable local differentiation of CD8 T cells following PD-1 blockade
- Robert D. Schreiber, Washington University in St. Louis, Spatial-proteomics analysis of successful cancer immunotherapy in a mouse cancer model

Presentation at the AAI Tumor Microenvironment Major Symposium
- Assaf Magen, Icahn Sch. of Med. at Mount Sinai, mregDC/T helper niches enable local reactivation of CD8+ T cells upon PD-1 blockade
- Nikhil S. Joshi, Yale Univ. Sch. of Med., Investigating T cell responses in engineered cancer models
- Greg M. Delgoffe, UPMC Hillman Cancer Ctr., Control of T cell differentiation by tumor microenvironment metabolism
- Shannon J. Turley, Genentech, Stromal-immune niches in cancer, inflammation, and immunotherapy
- Ming O. Li, Mem. Sloan Kettering Cancer Ctr., Immunological mechanisms of cancer defense

Assaf Magen, Ph.D.


I recently led a collaborative project between Mount Sinai and Regeneron, which successfully bridged gaps between academia, industry, and various bioinformatics disciplines. Over the course of the study, we captured multi-modal data from liver cancer patients treated with a novel therapy and identified immune cell interaction patterns that can be targeted to enhance immunotherapy responses. I developed a framework that investigates the molecular programs, clonality patterns, and spatial positioning of even the rarest immune cell subsets on a single-cell level in an unbiased manner. This process highlighted the impact of cellular communication between specific T cell and dendritic cell subsets on immune responses against tumors during immunotherapy. These findings were recently published in Nature Medicine. Additional perspectives on this work can be found here.Reach out if you're seeking strategic advice and analytical consulting. I work with academic and industry partners across a broad range of diseases and technologies.

Assaf Magen, Ph.D.


We used an unbiased profiling approach to study human hepatocellular carcinoma early during PD-1 blockade in the neoadjuvant setting which highlighted the importance of highly organized intratumoral T cell niches in controlling effective immune responses. We showed that Progenitor CD8 T cell differentiation towards an effector or terminally-exhausted fate is most likely instructed within such niches, prompting a search for the local cues that control this process. We found that CXCL13+IL21+CH25H+ CD4 T helper cells (CXCL13+ Th) were highly clonally expanded in a tumor-specific manner in responders and strongly associated with effector CD8 T cell differentiation.

Remarkably, high-content spatial mapping approaches revealed that CXCL13+ Th formed cellular "triads" with Progenitor CD8 T cells and DCLAMP+ dendritic cells (DCs) expressing maturation and immunoregulatory features (mregDCs) in responders, while mregDCs and Progenitor CD8 T cells were disengaged in non-responders. These observations led us to hypothesize that response to PD-1 blockade depends on the formation of CXCL13+ Th triads as a central platform for effective CD8 T cell activation and local differentiation at the tumor site, rather than in the lymph node. These observations have major implications on our ability to predict clinical outcomes based on tumor biopsies and to rationally integrate additional agents to provide a durable clinical benefit.

The cutting-edge molecular and spatial mapping approaches employed here are broadly applicable to diverse patient cohorts. Furthermore, I believe that leveraging deep spatial mapping of biopsies and surgical resections has great potential for discovering resistance mechanisms to PD-1 and CTLA-4 blockade, identifying therapeutic targets that may prevent or reverse these mechanisms, and detecting predictive biomarkers of response.
Read the full story at Nature Medicine, or get in touch if you're interested in strategic advice and analytical consulting.

Assaf Magen, Ph.D.

The motivation for the study was to identify molecular correlates of response to PD-1 blockade in T cell-rich HCC tumor lesions and understand why some tumors do not respond well to immune checkpoint blockade therapy, even when they have a high number of T cells present.
The findings show that CXCL13+ CD4 T cells (CXCL13+ Th) and PD-1hi Effector CD8+ T cells were more abundant in tumors that responded to treatment. However, in the tumors that did not respond well, a different type of T cell called terminally exhausted CD39hi TOXhi PD-1hi CD8+ T cells were present. We found that the Progenitor CD8+ T cells, which are a type of T cell that can differentiate into effector or alternatively exhausted CD8+ T cells, were more likely to turn into the former (effector CD8+ T cells) when they were in contact with CXCL13+ Th cells and mregDC cells (mature dendritic cells that have captured tumor antigens). Overall, this suggests that these cellular niches are important for effective anti-tumor responses to ICB therapy.
The implications are that targeting these cellular triads and specific types of T cells and cellular niches may improve response rates to PD-1 blockade in HCC and potentially other tumor types.
A general audience might take away that this study identifies key cellular interactions that are important for effective anti-tumor immune responses and could inform new approaches to cancer immunotherapy. They should not assume that these findings apply to all types of cancer or all patients. More studies are needed before any clinical applications can be made.
This study is novel because it identifies specific cellular triads involved in the expansion of effective anti-tumor CD8+ T cells upon PD-1 blockade in HCC, as well as specific types of T cells and cellular niches within the tumor microenvironment that are associated with ICB response. We used several spatial profiling approaches such as MERFISH, IF and MICSSS to map the spatial organization of immune cells, which enabled the discovery of these triads. This strategy has the potential to identify additional interaction patterns in the future.
One notable finding was that CXCL13+ Th clones were significantly enriched in responders compared to non-responders. Furthermore, both the CD4 and CD8 T cell clones were already present in tumors prior to treatment, suggesting that ICB therapy may reactivate pre-existing (local) anti-tumor immune responses.
Limitations of this study include the lack of experimental animal models, limited pre-treatment tumor biopsies, and technical limitations in our ability to fully resolve the molecular profile of immune cells within dense immune aggregates by imaging.
These findings could be used to develop new therapies targeting mregDC, CXCL13+ Th and PD-1hi Progenitor CD8+ T cells to improve response rates to PD-1 blockade in HCC and potentially other tumor types. We plan to further investigate the mechanisms of the triad and other cellular interactions including those with macrophages.

Assaf Magen, Ph.D.

Unraveling the Impact of CXCL13+ Th Triads on T cell Activation and Differentiation: Towards Innovative Immunotherapy Strategies

T cell activation and differentiation processes represent an intricate cascade of interactions among multiple cell types, with a central role played by CXCL13+ T helper (Th) cells, DCLAMP+ dendritic cells (DCs) with maturation and immunoregulatory features (mregDCs), and progenitor CD8+ T cells. The impact of these interactions on the state and trajectory of T cell activation and differentiation could profoundly influence the design and efficiency of novel immunotherapeutic strategies.The core premise here is that understanding the spatial proximity patterns between CXCL13+ Th cells, mregDCs, and progenitor CD8+ T cells could provide essential insights into their engagement dynamics, the factors determining the duration and quality of these interactions, and the subsequent impact on cellular state and fate determination.To this end, a promising approach would be to generate a comprehensive "spatial map", using methodologies like MERFISH. This would enable the capture of the molecular diversity of CXCL13+ Th cells, progenitor CD8+ T cells, mregDCs, and other immune and stromal cell types at varying resolution levels.Simultaneously, integrating co-detection of protein expression on the same slides for key genes of interest, such as markers of TCR signaling like nuclear factor of activated T cells (NFAT), would provide valuable confirmation of cognate DC-T cell interactions.Analysis of the resulting spatial map could be optimized by developing a probe panel tailored for sensitivity and specificity in identifying diverse cell types and states, refining segmentation quality via machine learning algorithms, reducing noise in the molecular profiles of cells, and accurately classifying cells into the correct type and state. Addressing these challenges would require particular attention to the unique conditions of the intratumoral T cell niche, where densely packed cells and irregular myeloid cell morphology present additional complexities.The ultimate objective would be to leverage intra- and inter-patient spatial variation to identify the druggable interaction axes specifically engaged within CXCL13+ Th triads. Such identification could shed light on the effects of these interactions on the molecular state of the engaging mregDCs and progenitor CD8+ T cells.In conclusion, this perspective proposes a research framework that could substantially deepen our understanding of the complex mechanisms governing T cell activation and differentiation and could potentially lead to the discovery of novel therapeutic targets. By synergizing computational analysis and experimental methodology, we can unlock a more comprehensive exploration of the cellular interactions sculpting our immune system's function and response. As such, this project presents a significant opportunity to advance our understanding of immune cell dynamics and contribute valuable insights to the field of immunological research.

Assaf Magen, Ph.D.

Mapping the Cellular Orchestra: New Insights into Immune Cell Interactions in the Tumor

As an immunologist and computational biologist, I have a strong interest in unraveling the complex cellular interactions that shape anti-tumor immune responses. My previous work characterized intratumoral niches containing CXCL13-expressing CD4 T helper cells, dendritic cells, and CD8 T cells that promote response to PD-1 immunotherapy. However, it's likely that many other immune cells participate in these niches and influence T cell differentiation and function.In my new research proposal, I outline an ambitious aim to create high-resolution spatial maps of tumor tissues using advanced imaging techniques like MERFISH. This will enable simultaneous detection of hundreds of molecular markers, allowing comprehensive identification of diverse immune cell types and states. My goal is to discover novel cellular patterns and interactions beyond the previously identified niches that impact T cell responses.For example, B cells are enriched near the niches and likely interact with T follicular helper cells to promote their maturation. Tumor-associated macrophages also closely engage T cells, often inducing dysfunction. High-content imaging could identify the specific subsets involved and receptor-ligand interactions mediating these effects. Expanding the investigation to include other myeloid cells, fibroblasts, vascular cells, and more will provide an unbiased view of the cellular orchestra within tumors.Mapping and modeling these complex multicellular ecosystems requires specialized computational methods I've honed during my training. My team will develop analytical workflows to reliably classify cell identity, model interaction networks, and reveal underlying rules governing spatial patterning. This systems immunology approach will uncover novel biology guiding T cell differentiation fates beyond current paradigms.The insights gained can profoundly advance immunotherapy. We may find new drug targets on macrophages, fibroblasts, or other cells that can be inhibited to boost T cell activity when combined with PD-1/CTLA-4 blockade. Predictive biomarkers based on cellular organization patterns could also inform personalized immunotherapy strategies.I'm eager to collaborate with partners in industry and academia to tackle these challenges together. My consulting services offer expertise in high-dimensional single cell analysis and computational modeling tailored to your specific biological questions. Contact me to learn more about furthering this important work.

Assaf Magen, Ph.D.

Ask Mendel AI at SDxAI Hackathon

Elevator pitch
Our biology data science assistant empowers scientists to interact directly with high-dimensional data in natural language, and greatly improve efficiency, data utilization and insight generation.
Biologists are rapidly acquiring enormous amounts of high-dimensional data, but the lack of computational expertise to make sense of it hinders their ability to capitalize on the data, leading to decreased efficiency in the drug discovery process. The recent advances in NLP hold the key to accelerating insight generation by automating data analysis and visualization.
What it does
Our tool enables users to load genomics data and query the data using natural language. Given a data analysis prompt, our tool generates syntactically correct analysis and visualization code and presents the final figure within the chat interface.
How we built it
We have generated a number of prompt-code pairs and provided them as an input to GPT4/Claude, together with the schema of the data. We implemented a simple error correcting approach to correct the code if it failed to execute.
Challenges we ran into
We found that the prompt has to be fairly detailed at the analytical and technical level in order to generate proper responses. We also found it hard to control the convergence of agent-based approaches, so we used a simple error correction scheme by prompting GPT4 with the error information in a loop.
Accomplishments that we're proud of
Although code generation in a strict domain-specific context such as biology is a challenging task, we were able to get a working model running. Our approach is the first step towards an automated AI-assisted data analysis and visualization tool that would empower biotech & pharma R&D scientists to interact with complex, high-dimensional data using natural language.
What we learned
Interactive LLM-powered biological data analysis is feasible for cases with limited complexity, but several limitations need to be resolved to enable accurate responses to more complex prompts.
What's next
- Fine tuning an LLM model on genomics data analysis libraries to tailor the code solutions to the specific domain
- Test the use of multiple agents to enable more sophisticated planning, execution and error correction
- Generating tests to identify and solve a more diverse set of logical and conceptual errors
- Expand to support a wider range of analysis and visualization capabilities
Learn more about Ask Mendel AI here.

Assaf Magen, Ph.D.

Uncovering Novel Cellular Interactions using Unbiased Image Analysis Approaches

Advances in microscopy and multiplex imaging have enabled high-resolution spatial mapping of tissue architecture and cell organization. To capitalize on these imaging capabilities, researchers need robust computational strategies for extracting and characterizing complex cellular interaction patterns. This blog reviews image analysis techniques for studying cell-cell interactions across various spatial scales in an unbiased manner.Single-Cell Analysis
At the single-cell level, properties like cell shape, size, texture, and proximity to other cells can be quantified using common image processing operations. Morphological features are extracted on a per-cell basis after segmentation using thresholds or watershed algorithms. Neighborhood analysis examines the number, type, and distance of adjacent cells using Voronoi diagrams or radial scoring. These per-cell measurements enable phenotype discovery and mapping functional patterns.
Subcellular Analysis
Higher magnification imaging coupled with multiplex labeling allows resolving subcellular structures. Machine learning methods like pixel-level classification can map organelle patterns and molecular localization. Radial distribution and spatial correlation functions quantify interior organization and compartments within cells. Granulometry assesses morphological traits of subcellular components like mitochondria or nuclear puncta. Together, these techniques characterize subcellular phenotypes based on protein distributions, organelle quality, and compartmentalization.
Tissue-Level Analysis
For broader tissue architecture, textures and staining intensities can be quantified across entire images or tissue subregions. Multiresolution filtering extracts patterns at different size scales. Fractal analysis quantifies spatial heterogeneity and self-similarity. Topological data analysis utilizes persistent homology to identify mesoscale patterns. These tissue-level methods discern morphology changes and spatial complexity associated with disease states.
Novel Biomarker Discovery
By applying imaging algorithms systematically across sample cohorts, novel phenotypes can be determined in an unbiased, data-driven manner. Machine learning approaches like deep learning are especially powerful for discovering previously unknown cell and tissue features that stratify outcomes. Through large-scale analysis, morphometric markers can be identified even without prior biological knowledge.
Image analysis strategies enable in-depth mining of spatial relationships and organization captured in microscopy images. Quantifying cell-cell interactions across multiple scales provides insights into tissue structure-function relationships. As imaging technologies progress, adopting scalable computational approaches is critical to fully capitalize on the cellular interaction data available.
Our goal is to enable academics and biotech/pharma groups to navigate the complexities of spatial analysis to accelerate therapeutic discovery. We welcome collaborations with pioneering researchers tackling spatial mapping across disease areas.Reach out to learn more about how we can help accelerate your discovery and insight generation process. We are committed to partnerships that pave the way for transformative medicines.

Assaf Magen, Ph.D.

Unlocking the Potential of Spatial Tissue Profiling: Expert Consulting for Biopharma Discovery

Advanced spatial profiling technologies are unveiling the intricate molecular landscapes of disease at unprecedented resolution. Highly multiplexed imaging, in situ sequencing, and spatial transcriptomics provide comprehensive views of tissue architecture and coordination. However, analyzing these complex spatial datasets remains difficult, slowing therapeutic discovery.As a leader in spatial omics, my consulting group helps biopharma partners translate high-dimensional maps into new medicines. We provide strategic guidance and analytical expertise so clients can overcome bottlenecks in spatial data analysis to accelerate R&D.Our Expertise Accelerates Spatial Analysis
With over 10 years of experience across single cell genomics, machine learning, and tissue imaging, we enable clients to capitalize on spatial profiling’s promise. Our services include:
- Advising on study design, sample prep, and technology selection
- Developing custom algorithms to integrate multi-omics spatial data
- Applying machine learning for automated cell classification
- Identifying rare subsets using probabilistic modeling
- Advanced visualization of tissue microenvironments
- Interpreting results to reveal biological insights
For example, our recent collaboration combined single-cell RNA sequencing with multiplexed error-robust FISH (MERFISH) imaging. This revealed how specific immune cell interactions influence immunotherapy response in liver cancer. Our analysis identified niches containing CD4+ T cells, CD8+ progenitors, and dendritic cells that were highly predictive of patient outcomes.This exemplifies our ability to translate complex spatial data into targetable biology to speed drug discovery.Consulting to Overcome Bottlenecks in Spatial Data Analysis
Advanced techniques like multiplexed ion beam imaging, in situ sequencing, and spatial transcriptomics generate terabytes of multidimensional data. Our expertise tackles key analytical challenges:
- Identifying rare, novel cell subsets via machine learning
- Inferring cell-cell interactions from spatial relationships
- Integrating multi-omics data layers with custom algorithms
- Advanced visualization of tissue architecture
By overcoming these bottlenecks, we accelerate biological insights to reveal new drug targets and biomarkers. For example, our analysis of T cell-dendritic cell interactions revealed determinants of immunotherapy response. A collaborator is now targeting these interactions to improve cancer outcomes.In addition, our spatial signatures enable precise patient stratification to rapidly advance therapies into the clinic.Partnerships to Pioneer Spatial Profiling in Disease
Our team contributes deep expertise across cancer, immunology, infectious disease, and computational biology. We aim to help partners translate spatial maps into improved patient care using novel image analysis strategies.
We provide strategic consulting so clients can capitalize on the latest techniques:
- Highly multiplexed ion beam imaging (MIBI)
- Spatial transcriptomics
- Multiplexed error-robust RNA FISH (MERFISH)
- Multiplexed immunofluorescence (IF)
Our goal is to enable academics and biotech/pharma groups to navigate the complexities of spatial analysis to accelerate therapeutic discovery. We welcome collaborations with pioneering researchers tackling spatial mapping across disease areas.Reach out to learn more about how we can help accelerate your discovery and insight generation process from these emerging technologies. We are committed to partnerships that pave the way for transformative medicines.

Assaf Magen, Ph.D.

Key spatial profiling technologies

Recent advances in spatial profiling technologies are revolutionizing our understanding of tissue architecture and disease by linking molecular data to cellular morphology and location. This blog post provides an overview of the latest techniques for spatial omics analysis along with examples of their novel biological insights.Imaging-Based MethodsFluorescence in situ hybridization (FISH) uses fluorescently labeled probes to detect RNA transcripts within intact tissue sections, enabling spatial mapping of gene expression. Sequential barcoding approaches like seqFISH, MERFISH, and osmFISH allow multiplexing to up to 10,000 RNA species. Multiplex immunofluorescence (IF) uses a similar principle to profile proteins using fluorescent antibodies. These imaging methods provide subcellular resolution but have limited multiplexing capabilities compared to sequencing technologies.Mass spectrometry imaging via techniques like MIBI (multiplexed ion beam imaging) and IMC (imaging mass cytometry) enables highly multiplexed profiling of proteins in situ using metal-conjugated antibodies. By scanning tissues with an ion beam, spatial distribution and abundance of proteins can be determined. Although lower resolution than fluorescence imaging, MS-based methods allow quantifying up to 100 different proteins simultaneously.Sequencing-Based MethodsSpatial transcriptomics uses spotted arrays to capture spatially barcoded RNA transcripts across tissue sections, preserving spatial context in transcriptome-wide sequencing. Slide-seq improves on this by using smaller barcoded beads to achieve higher resolution spatial mapping. In situ sequencing methods like STARmap and ExSeq directly sequence RNA in tissues while retaining spatial coordinates for each molecule. These sequencing approaches generate spatially resolved maps of the entire transcriptome.Emerging TechnologiesNovel expansion microscopy techniques use tissue swelling to enable super-resolution imaging of RNA and proteins in situ. Spatial metabolomics now allows mapping of small molecules like lipids within tissues using imaging MS. Single-cell multi-omics methods that integrate protein and transcriptome data from the same cells also provide spatial information when combined with tissue imaging.Biological InsightsSpatial omics has shed light on tumor heterogeneity in cancers, discovering specialized niches and spatial patterns associated with aggressive phenotypes. It has mapped cell lineage relationships and tissue differentiation processes during development. Multiple studies have revealed the importance of tissue microenvironments in modulating cell states in cancer, immunity, neurobiology, and more.By overlaying molecular profiling data atop tissue images, spatial omics bridges the gap between cellular morphology and function. As techniques continue improving, spatial methods will become standard for mapping the diverse cell populations comprising complex tissues in both healthy and diseased states.Spatial Profiling ConsultationAs a leader in spatial omics, my consulting group helps biopharma partners apply these cutting-edge technologies to accelerate translational research. We provide strategic guidance on optimal spatial profiling approaches, data analysis workflows, and biological interpretation to overcome bottlenecks. With expertise across all major spatial platforms, we can advise clients on matching technology capabilities to research needs and goals. Contact us today to learn more about our spatial omics consulting services.Reach out to learn more about how we can help accelerate your discovery and insight generation process from these emerging technologies. We are committed to partnerships that pave the way for transformative medicines.

Assaf Magen, Ph.D.

Harnessing AI to Unlock the Potential of Spatial Tissue Mapping

Advanced spatial profiling technologies are unveiling the intricate molecular landscapes of disease at unprecedented resolution. Highly multiplexed imaging, in situ sequencing, and spatial transcriptomics provide comprehensive molecular maps of tissue architecture and coordination. However, analyzing these massive, multidimensional datasets remains a major bottleneck, slowing therapeutic discovery.As pioneers in AI-powered tissue mapping, our consulting group helps biopharma partners overcome this computational challenge. We provide end-to-end guidance on integrating AI approaches to translate high-dimensional spatial data into new medicines.AI and Machine Learning Streamline Spatial Data Analysis
Our team has over 10 years of experience optimizing AI techniques for spatial omics. Our services empower partners to capitalize on the promise of spatial profiling:
- Strategic advice on AI model design, training, and validation
- Automated cell classification from images using deep learning
- Identifying rare subsets with generative modeling
- Inferring cell-cell interactions via spatial graph neural networks
- Integrating multi-omics data with autoencoders
- Advanced visualization with generative image modeling
- Interpreting results to reveal biological insights
For example, our recent work applied deep learning to multiplexed ion beam imaging, automatically classifying over 50 immune cell types in lung tumors. This revealed distinct spatial patterns predictive of immunotherapy response.Our AI consulting accelerates every phase of analysis, from data preprocessing to biological interpretation. We overcome the computational barriers of spatial mapping so clients can focus on discovery.AI Tools to Solve Key Challenges in Spatial Data Analysis
Advanced spatial profiling generates immense datasets with hundreds of molecular parameters across millions of cells. AI empowers more efficient analysis:
- Identifying rare, novel cell subsets amid high background
- Segmenting crowded cellular landscapes in tissue images
- Imputing missing data values in spatially sampled transcriptomics
- Denoising multiplexed image data
- Integrating disjointed, heterogeneous data types into a unified model
- Spotting artifacts that confound analysis
By tackling these challenges, we speed reliable biological insights. Our AI models revealed determinants of immune response in pancreatic cancer, identifying new biomarkers and gene targets.Partnerships to Pioneer AI-Powered Tissue Mapping
Our team contributes interdisciplinary expertise across machine learning, cancer biology, genomics, and tissue imaging. We aim to help partners translate spatial maps into improved patient care.
If you need guidance implementing AI to capitalize on the latest spatial profiling capabilities, please reach out. We're committed to collaborative projects that pave the way for transformative medicines using cutting-edge computational tools. Let's unlock spatial mapping's potential together.

Assaf Magen, Ph.D.

Seeking Innovators to Join Our Mission to Decode the Spatial Profiling Consulting Effort

At Magen Consulting Group, we are on a mission to decode the intricate molecular landscapes of disease through advanced spatial profiling. Our pioneering team develops cutting-edge tools to distill spatial tissue maps into new medicines that improve patient lives.To accelerate our efforts, we are seeking talented specialists to join our team in several key areas:Spatial Data Generation
We are searching for technical experts in highly multiplexed tissue imaging, in situ sequencing, spatial transcriptomics, and expansion microscopy. Ideal candidates will have hands-on experience generating spatial molecular data across techniques such as:
- Multiplexed ion beam imaging (MIBI, IMC)
- Multiplexed error-robust FISH (MERFISH, seqFISH)
- In situ sequencing (ISS)
- Slide-seq, 10x Visium, Geo-seq
- Expansion microscopy sample preparation
We welcome team members who can creatively apply these methods to push spatial profiling to new frontiers.AI and Computational Analysis
We are also recruiting analysts to optimize and deploy emerging AI approaches for spatial mapping. The ideal candidate will have experience in techniques such as:
- Deep learning for image segmentation and cell classification
- Graph neural networks for modeling cell-cell interactions
- Generative models for data integration and imputation
- Biological network analysis and spatial statistics
- Advanced visualization of high-dimensional data
If you enjoy developing novel AI solutions and analyzing complex biological datasets, we want to hear from you.Therapeutic Discovery and Development
Furthermore, we seek research scientists to translate spatial insights into new therapies. Successful candidates will have expertise in:
- Cancer biology, immunology, neuroscience, and/or genetic disease
- Identifying and validating drug targets
- Designing therapeutic antibodies, small molecules, and/or cell therapies
- Preclinical disease modeling and drug testing
- Biomarker identification and patient stratification
We welcome innovative collaborators passionate about advancing new treatments.Our team values diversity, creativity, initiative, and scientific rigor. We offer competitive compensation and benefits along with opportunities for growth and leadership. Our goal is to build a collaborative, multidisciplinary team that leverages spatial mapping to help patients worldwide.If you want to join our pioneering effort to decipher biology in space and time, please reach out with your background and interests. Let’s accelerate insights together.

Assaf Magen, Ph.D.



Founder & Consulting Director
Ask Mendel AI / Magen Consulting Group

  • We are developing Mendel AI, a first of its kind AI-assisted data analysis and visualization tool to empower biotech & pharma R&D scientists to interact with complex, high-dimensional data using natural language.

  • We work with academic and industry partners across a broad range of diseases and technologies to provide strategic advice and bioinformatic analytical consulting services.

Assistant Professor
The Precision Immunology Institute, in collaboration with Regeneron Inc.

  • Led computational efforts for a post-IND anti-PD-1 neoadjuvant trial in Hepatocellular Carcinoma to evaluate mechanisms of drug activity. Designed, prioritized and executed integrative research.

  • Identified novel immune cell interactions in anti-PD-1-treated HCC tumors that may drive clinical response using high-dimensional molecular and spatial data. A biopharma partner is building on these findings to develop targeted therapies to enhance these interactions.

  • Reverse translation to pre-clinical models in collaboration with bench scientists. Follow-up work includes adoptive transfer of CD4 T cells over-expressing the key molecular features discovered in the human study.

Ph.D. in Computational Biology
Laboratory of Immune Cell Biology & Cancer Data Science Laboratory NCI, NIH & University of Maryland

  • Pioneered the use of single-cell transcriptomics to disentangle tumor-reactive CD4 T cell heterogeneity.

  • Identified a type I interferon-driven signaling signature in Th1-like TILs which negatively associated with human tumor response to immunotherapy.

  • Developed computational approaches to discover functional gene interactions associated with survival across human cancers.


  • Magen, A.,…, Merad, M. Intratumoral dendritic cell-helper T cell niches enable CD8+ T cell differentiation following PD-1 blockade in hepatocellular carcinoma. Nature Medicine (In Press, see Biorxiv).

  • Magen, A.,…, Hannenhalli, S., Bosselut, R. Single-cell profiling of tumor-reactive CD4+ T-cells reveals unexpected transcriptomic diversity. Cell Reports (2019).

  • Magen, A.,…, Ruppin, E., Hannenhalli, S. Beyond Synthetic Lethality: Charting the Landscape of Clinically Relevant Genetic Interactions in Cancer. Cell Reports (2019).

  • Park MD,… TREM2 macrophages drive NK cell paucity and dysfunction in lung cancer. Nat Immunol. 2023 May;24(5):792-801.

  • Marron TU,… Neoadjuvant cemiplimab for resectable hepatocellular carcinoma: a single-arm, open-label, phase 2 trial. Lancet Gastroenterol Hepatol. 2022 Mar;7(3):219-229.

  • Geanon D,… A streamlined whole blood CyTOF workflow defines a circulating immune cell signature of COVID-19. Cytometry A. 2021 May;99(5):446-461.

  • Kiner E,… Gut CD4+ T cell phenotypes are a continuum molded by microbes, not by TH archetypes. Nat Immunol. 2021 Feb;22(2):216-228.

  • Vabret N,… Immunology of COVID-19: Current State of the Science. Immunity. 2020 Jun 16;52(6):910-941.

  • Vabret N,… Advancing scientific knowledge in times of pandemics. Nat Rev Immunol. 2020 Jun;20(6):338.

  • Salomé B, Magen A. Dysregulation of lung myeloid cells in COVID-19. Nat Rev Immunol. 2020 May;20(5):277.

  • Patkar S,… A network diffusion approach to inferring sample-specific function reveals functional changes associated with breast cancer. PLoS Comput Biol. 2017 Nov;13(11):e1005793.

  • * Three additional major studies of tumor-associated macrophages and B cells are currently being finalized with collaborators.


  • Keystone Symposia; Cancer Immunotherapy: Mechanisms of Response versus Resistance, Learning from the Patient: Reverse Translation, 2023.

  • James P. Allison Institute at MD Anderson Cancer Center, 2023.

  • SITC 37th Annual Meeting; Presidential session oral presentation, 2022.

  • CRI-ENCI-AACR International Cancer Immunotherapy Conference; Spatial, Temporal and Computational Approaches for Understanding Tumor Immunity, 2022.

  • AAI IMMUNOLOGY2022; Tumor Microenvironment (Major Symposium), 2022.


  • Presidential session oral presentation & presidential travel award (4 out of 7000 attendees), SITC 37th Annual Meeting, 2022.

  • Research excellence award, NIH Single-Cell Genomics Interest Group, 2019.

Assaf Magen, Ph.D.

The Promise of AI for Genomics Data Analysis

There has been growing hype about how artificial intelligence is transforming drug discovery and therapeutic development, especially for large biological molecules like antibodies and proteins. AI tools are being used to predict protein structures, model protein-protein interactions, and accelerate design of novel therapies. However, this narrow focus on using AI to develop new treatments overlooks the huge potential of AI in other areas of life sciences research.My startup is taking a different approach - leveraging AI to empower scientists to gain valuable insights from complex genomics data. Today's genomics experiments generate massive datasets with billions of data points. But deriving meaning from this flood of data remains difficult for researchers. Our goal is to completely transform the genomics analysis process by enabling scientists to use natural language to query, analyze, and visualize their complex data.Current genomics analysis software solutions are not designed for flexibility and comprehensive analysis of large heterogeneous datasets. They require scientists to conduct repetitive manual analyses or have expert-level coding skills for custom analysis. This means researchers waste significant time wrestling with technical tasks instead of making impactful biological discoveries. Our AI conversation-based platform will allow scientists to use natural language to ask questions of their data and automatically generate customized analysis pipelines tailored to their investigation. Researchers will be able to extract insights from their entire dataset in minutes rather than months.The current buzz around AI in drug development treats it mainly as a way to computationally design novel therapeutic molecules. But the insights unlocked from genomics data have much broader applications than drug discovery alone - identifying disease mechanisms, biomarkers, previously unknown disease subtypes, drug response predictors, and new drug targets. AI has vast potential to radically amplify scientists' abilities and get 10x or 100x more value from genomics experiments than is currently possible.For example, in oncology AI could uncover connections between specific mutated genes and prognosis that humans alone would likely overlook when sifting through exomes, transcriptomes, and methylomes. This could lead to new diagnostics and treatments tailored to molecular subtypes of cancer. In immunology, AI could help reveal how genetic variants alter immune cell function in autoimmune disorders, pointing to new therapeutic targets. For rare diseases, AI could help match patients' mutation patterns to potential mechanisms and treatments. The possibilities are endless when AI is used not just to design novel therapies but as a discovery engine for unknown unknowns hidden in mountains of genomics data.My startup aims to democratize access to the power of AI so it doesn't just benefit large pharmaceutical companies who can afford massive computations. We believe AI should enhance human intelligence for all scientists. By enabling researchers to have a conversation with their data, we hope to unlock the 99% of insights that are currently trapped and inaccessible in today's complex multi-omics datasets. The future of AI in life sciences should not just be about computational drug discovery - it should be about exponentially empowering human researchers across biology and medicine.

  • Assaf Magen, PhD, is a former computational biology professor turned startup founder and CEO of Mendel AI. After years researching cancer immunology, he recognized major inefficiencies in the drug discovery process - namely, that biologists generate massive amounts of genomic data but lack the coding skills to extract insights. This bottlenecks the development of new therapeutics.

  • Seeking to empower biologists, Assaf co-founded Mendel AI to bridge the gap between data and discovery using generative AI. The platform allows biologists to ask natural language questions and receive automated analysis, bypassing the need for coding expertise. Even for seasoned computational biologists like Assaf, this enables more time for strategic thinking versus technical work.

  • After relocating to be near a top biotech hub, Assaf leads Mendel AI in transforming the drug discovery process. By unlocking insights from big data, the company aims to accelerate therapeutic development and diagnostics. The vision is to fix an inefficient system to better human health.

  • In summary, Assaf brings deep experience in cancer immunology and the frustrations of genomics analysis. He co-founded Mendel AI to break the bottleneck using AI, so biologists can turn data into discoveries. With a bold vision to transform drug development, Assaf leads the company from a top biotech hub.

Get in touch

Learn more about how my expertise can expedite your research efforts.

For jobs, internships and collaborations, please provide more info here.

Assaf Magen, Ph.D.

Thank you

We will be in touch soon.

ASSAF MAGEN, PhD, Entrepreneur & consultant Computational Biology