There’s a common thread across every industry: scientific revolutions are sparked by innovations in measurement technology. This stems from the fundamental truth that we can only optimize what we can measure.
Drug discovery is no different. The goal is to optimize healthy tissue function by altering the molecular states and interactions of specific cells. But connecting the function of cells in tissues to their underlying molecular states has been a major challenge, making drug discovery an immensely difficult optimization problem.
Over nearly a decade, our founding team has serially invented frontier technologies to overcome this measurement problem (including SABER-FISH and Light-Seq), culminating in a scalable solution that unifies the most powerful measurements in biology today. We’ve industrialized our innovations into the M×P Platform (Molecules × Phenotypes) – a next-generation measurement engine that integrates high-resolution imaging with next-generation sequencing (NGS), computation, and AI lab automation to connect cells in tissues to their underlying molecules, with breakthrough precision and scale.
We founded Digital Biology to unite world class scientists, engineers from top technology companies, and seasoned drug hunters on a quest toward a singular goal: bring precise and scalable tissue measurements to every stage of discovery, engineering next-generation therapies with optimal function in human tissues.
We’ve partnered with world-leading pharma companies and global foundations to develop and scale this platform. Now, Digital Biology is applying this paradigm shift in measurement to engineer precision biologics for solid tumors. The M×P Platform drives three core pillars of our discovery engine:
This piece dives deep into a case study in Target Discovery: identifying pairs of spatially synergistic targets that unlock precise and effective anti-tumor immunotherapies. These insights are a launch point for Digital Biology’s internal programs, compounding the value of clinically validated targets in precision oncology.
In the coming months, we’ll reveal how measurement-driven innovation powers our full-stack biologics discovery engine—from target discovery, to sequence-to-function screening in human tissues, and ultimately, to the next generation of optimized immunotherapies with deep characterization in human tissues.
Scalable spatial sequencing in fixed tissues, guided by digital histology.
At Digital Biology, we are engineering precision biologics that tackle key challenges in solid tumor treatment. Our goal is to enhance on-tumor specificity and efficacy by developing biologics and combination therapies that harness spatially synergistic anti-tumor effects.
A major challenge in solid tumor targeting is the heterogeneous spatial distribution of tumor antigens, which limits the effectiveness of T cell engagers, CAR-T therapies, and ADCs. Traditional therapies often fail to address antigen loss and heterogeneity, leaving tumor cells untargeted and increasing relapse risk. However, in this challenge lies an opportunity.
By mapping and exploiting the spatial organization of tumor antigen expression, we're discovering multi-targeting strategies that leverage complementary and/or overlapping spatial localization of targets to achieve enhanced specificity, efficacy, and durability of response.
Recent success of spatially synergistic immunotherapies, such as bispecific antibodies targeting PD-1 and VEGF, demonstrate the power of this approach, but breakthroughs have been limited by a lack of scalable technologies to discover spatially optimal target pairs. Our platform enables discovery and validation of spatially synergistic targets within the tumor microenvironment (TME).
Building on clinically validated tumor-associated antigens (TAAs), we are discovering secondary targets that:
A set of promising, clinically validated TAAs were prioritized for co-target discovery based on existing clinical data, high prevalence in non-small cell lung cancer (NSCLC) patient tumors, and fit for our therapeutic hypotheses. Our goal was to discover secondary targets with co-localized or complementary expression within patient tumors, informing design of multi-targeting and combination therapies, respectively. Here, we present a preview of results for CEACAM5, a widely studied and clinically validated solid tumor antigen.
Patient tumors were screened for CEACAM5 protein expression
Tumors expressing CEACAM5
were selected for spatial sequencing
Sequencing of cells with high CEACAM5 surface protein expression to identify spatially correlated targets
Sequencing of cells with low CEACAM5 surface protein expression to identify spatially complementary targets
More than 50 patients were screened for CEACAM5 protein expression via immunofluorescence using a clinically validated antibody. Ten patient tumors that displayed significant CEACAM5 surface protein were selected for deep molecular investigation, and a total of >50 FFPE tumor sections (5 µm) were analyzed on the M×P Platform.
Using high-resolution protein images, AI models were trained to identify individual tumor cells and assign them a targetable_score, reflecting the level of CEACAM5 protein expressed on the cell surface. As CEACAM5 is known to localize to the cell membrane and cytosol, this is an important filter to hone in on cells that are potentially targetable by a CEACAM5-targeting biologic.
Based on this score, cells were classified into two groups: target_high and target_low. Target_high cells exhibit high surface-localized CEACAM5 protein, making them prime cellular targets for targeted therapies, while target_low cells display low surface-localized CEACAM5 protein, representing tumor cells which may evade CEACAM5-based targeting.
To identify spatially synergistic and complementary targets, both target_high and target_low tumor cell populations were subjected to spatial next-generation sequencing. This allowed us to identify tumor-associated antigens (TAAs) that either intersect with or complement CEACAM5 protein expression.
Tumors were screened for CEACAM5 protein expression by immunofluorescence
Tumors with spatial heterogeneity in CEACAM5 protein expression were selected for spatial NGS
High resolution protein imaging allowed single-cell digitization of tumors
AI models classify tumor cells into target_high and target_low cell populations
target_high and target_low cell classifications were used to guide spatial NGS in the tumor tissue
Barcoded transcriptomes are collected from the tissue and processed with next-generation sequencing, leaving the tissue section intact.
Spatial NGS discovered molecular underpinnings of CEACAM5 heterogeneity within and among patient tumors.
First, we verified that the spatial NGS libraries displayed high library quality. In total, 19,192 genes were detected, with capture of up to an average of 4,500 unique mRNA measurements per cell. Importantly, these libraries originated from cell populations of only 100s-1000s of cells, demonstrating spatial NGS from rare populations within FFPE tissue sections.
By principal component analysis, spatial NGS libraries cluster by patient, and further cluster according to pathologist-annotated NSCLC subtype. Sequenced cell populations displayed a wide range of CEACAM5 surface protein expression, quantified as the targetable_score.
The M×P Platform enables nucleotide-resolution spatial sequencing of TME cells, guided by targetable protein expression. This allows for differential gene expression analysis, isoform detection, and SNP identification in rare cancer cell populations—all while preserving spatial context.
In each patient, we compared CEACAM5-high (target_high) and CEACAM5-low (target_low) cellular transcriptomes. Despite their spatial intermingling, these populations exhibited distinct molecular profiles, with CEACAM5 mRNA significantly enriched in target_high cells. Volcano plots and heatmaps below highlight hundreds of differentially expressed genes, ranked by expression enrichment and statistical significance.
Beyond gene expression, the M×P Platform identifies nucleotide-level variants which can inform patient selection for targeted treatments, an example of which is shown in the missense mutation detected in CEACAM5 in Tumor 001. Across the transcriptome, we uncovered many potential genetic variants of functional interest, including predicted variants in EGFR:
While variant analysis was not the study’s focus, the M×P Platform offers the unique integration of spatial, genetic, transcriptomic, and proteomic data -- delivering a comprehensive molecular view of patient tumors. This technology enables target discovery, prioritization, and biomarker analysis, unlocking new spatially synergistic therapeutic opportunities.
Notably, many CEACAM5-correlated genes were shared across patients, suggesting conserved gene modules underlying tumor antigen heterogeneity that could guide spatially informed drug discovery.
Spatial NGS Data
Spatial NGS Data
Spatial NGS Data
To systematically identify spatially co-expressed therapeutic targets, we analyzed the full transcriptome to pinpoint genes whose expression significantly correlated or anti-correlated with CEACAM5 protein levels across patients. This analysis identified more than 1,000 genes in each category. Notably, pathway enrichment analysis revealed that tumor cells with low CEACAM5 expression exhibited an upregulation of genes linked to proliferation and metastasis, underscoring the biological significance of these alternative targets.
Among all genes that were detected in patient tumors, we identified 230 genes encoding targets of FDA approved therapies, 940 genes predicted to encode surface-localized proteins, and within this subset, 195 existing TAAs with clinical validation. Leveraging our M×P Platform, we integrated these findings with a vast dataset of more than 1 million healthy human cells, applying machine learning models to predict off-target risks, subcellular protein localization, antibody developability, and other proprietary filters. This target prioritization pipeline led to the identification of novel candidate TAAs and clinically validated TAAs with established therapeutic potential. Of these, many were significantly differential among CEACAM5 high and low cells in the TME, highlighted in the volcano plot below.
For both existing and novel TAAs, validation of protein expression and, critically, cell surface co-localization with CEACAM5 protein is a key criterion for prioritizing target pairs. To confirm localization at the protein level, several putative TAAs were selected for high-resolution imaging validation, based on their ranking in our target prioritization models.
Below, two TAA candidates are highlighted that were predicted to exhibit strong spatial synergy with CEACAM5 in our M×P Platform data: one target that co-localizes with CEACAM5 protein, and one which is located in complementary cells in the TME. These validation images were captured in patient tumors sequenced in this study, with representative images from Tumor 001 shown. Additional validation was conducted across more than 60 patient tumors, reinforcing the robustness of these high-potential, spatially synergistic targets.
Target selection remains one of the greatest challenges in developing effective solid tumor therapies. The problem is twofold: targets specific to the tumor microenvironment are often not expressed broadly or highly enough to drive robust responses, while more abundant targets tend to lack specificity, posing safety risks due to off-tumor expression.
A powerful path forward lies in identifying targets that work in spatial synergy -- complementary expression patterns that, when targeted together, enhance tumor specificity and coverage. This multi-targeting strategy holds the potential to overcome the limitations of single-target approaches, improving both efficacy and safety.
At Digital Biology, we're building large-scale patient datasets to systematically discover these spatially synergistic TAA pairs. Our platform enables rapid, data-driven co-target discovery across patients and validated targets. In just two weeks, we more than doubled the number of patients in our CEACAM5 dataset, only a glimpse of the proprietary data advantage we’re building.
Stay tuned -- we’ll soon share how our measurement-first approach is redefining which immunotherapy functions can be optimized at the very earliest stages of discovery, and how scalable molecular measurements unlock unprecedented insights into immunotherapy mechanisms, bringing a new level of precision to how these therapies are designed, tested, and refined.
We’re seeking clinical and strategic partners to leverage spatio-molecular insights across precision oncology at all stages of drug discovery and development.
