How spatial biology improves clinical trial success in oncology

Drug development in oncology is a dynamic and complex field. New therapeutic modalities appear constantly, ranging from direct tumor-targeting compounds to immunotherapies designed to re-engineer the microenvironment, aiming, for instance, to transform so-called “cold” tumors, that is, tumors that will likely not trigger a strong immune response, into “hot” tissues infiltrated by active immune cells and therefore more likely to respond to therapy.

However, despite these advances, the failure rate remains remarkably high¹. Many drugs show incredible promise in the lab, only to fail in Phase II clinical trials due to a lack of efficacy. Some of these failures can be attributed to the disconnect between macroscopic evaluation and microscopic reality. Assessment of response often relies on imaging protocols like RECIST², which measures success based solely on tumor shrinkage. Even with newer standards like iRECIST, which account for the initial pseudoprogression (swelling) caused by immune cell infiltration, we remain blind to the internal reality. These methods measure volume, not mechanism. A low-resolution volumetric scan cannot tell you if the tumor microenvironment is switching from an immunosuppressive state to an immune-activated one, nor does it capture the baseline architecture: identifying which patients had the specific spatial arrangement required for the treatment to work in the first place. Additionally, it also fails to capture the baseline architecture: identifying which patients had the specific spatial arrangement required for the treatment to work in the first place.

Enter spatial biology

To bridge this gap between macroscopic observation and microscopic reality, we need tools that can accurately resolve tissue architecture. This is where spatial biology comes in.

This rapidly evolving field encompasses a suite of technologies designed³ to map gene expression or protein abundance directly within the tissue. While traditional bulk sequencing yields a global expression profile of cell mixtures, and single-cell sequencing resolves detailed cell states at the cost of their spatial arrangement, spatial biology bridges the gap between the visual context of microscopy and the deep molecular profiling of omics: it allows scientists to see not only which cells are present, but where, and more importantly, how they interact with each other.

Lessons from the past

To highlight the potential value of spatial biology in clinical trials, we showcase two real-world examples where promising compounds failed because the intricate interplay between cancer cells and their microenvironment was likely overlooked.

In a first-in-human study on patients with solid tumors⁴, Roche evaluated a bispecific antibody designed to act as a molecular bridge, simultaneously targeting FAP on fibroblasts and DR5 on tumor cells, to trigger apoptosis⁵. While patients were selected based on the expression of both targets, the trial was discontinued due to a lack of clinical efficacy. The failure was likely not biochemical, but spatial: since the drug was designed to act as a cross-linker, it required cancer cells and fibroblasts to be located in immediate proximity⁶. In other words, the drug was reliant on a specific spatial niche where these cell types intermix. For tumors where cancer cells form a dense core with fibroblasts segregated at the edges, the drug would likely fail due to its inability to cross-link a sufficient number of DR5 receptors to trigger apoptosis. In such a context, spatial transcriptomics or proteomics can help distinguish between an ineffective drug and an effective but inaccessible target. This distinction is particularly critical for bispecific antibodies, which rely on the physical proximity of two distinct targets to function. This ensures that promising therapies are not prematurely discarded, but are instead stratified toward patients with the appropriate tissue architecture.

In a relatively similar setup⁷, Pfizer tested a bispecific antibody targeting both CD3 and P-Cadherin, a protein highly upregulated in various solid tumors⁸. Just like the Roche candidate, this drug acted as a molecular bridge, but leveraged a specific mechanism: directly recruiting T-cells to kill cancer cells. One arm binds to P-Cadherin on the tumor, and the other recruits circulating T-cells via CD3. This strategy was designed to bypass the need for natural antigen recognition, allowing the drug to artificially force any T-cell into an immunological synapse with the cancer cell. Despite promising preclinical data, the trial was terminated due to a lack of activity. As in the previous example, the spatial architecture of the tumor was overlooked. High-resolution spatial biology could have resolved the critical question of proximity: measuring the precise distance between T-cells and P-Cadherin positive tumor cells to verify if the “synthetic bridge” was actually formed. Furthermore, by comparing pre- and post-treatment biopsies, researchers could have determined if the drug successfully drove T-cells into the tumor core or if they remained excluded at the periphery, explaining the lack of clinical response.

From retrospective analysis to proactive strategy

Fortunately, facing these challenges, the industry is gradually incorporating spatial biology to guide clinical success. Recent breakthroughs demonstrate how this shift is actively refining patient stratification and de-risking early-stage development.

In a landmark 2025 study⁹ on non-small cell lung cancer (NSCLC), researchers identified distinct spatial signatures that accurately predicted immunotherapy outcomes where traditional biomarkers failed. Crucially, they found that resistance was driven by granulocyte-rich niches within the tumor compartment, while response relied on immune interactions rooted in the stromal compartment, a distinction impossible to see with bulk sequencing. Similarly, while tertiary lymphoid structures are widely recognized as strong predictors of immunotherapy response, their presence alone does not guarantee success. By digitally dissecting these structures using spatial transcriptomics, researchers identified specific fibroblast subtypes surrounding these immune hubs, creating an immunosuppressive niche that effectively prevents T-cells from attacking the tumor¹⁰.

While the previous examples illustrate the value of retrospective analysis, spatial biology is uniquely positioned to validate human efficacy prior to phase I, preventing costly downstream failures. In a pioneering phase 0 study¹¹, researchers injected micro-doses of a candidate drug directly into patient tumors before surgical resection and then employed spatial profiling to compare drug-exposed areas against unexposed neighboring regions within the same patient. The data proved that the drug successfully engaged its target and triggered the intended immune response in living human tissue, providing the necessary evidence to advance to phase 1.

While the potential is clear, implementing spatial workflows in a clinical biology setting requires addressing practical hurdles:

• • Scalability: Moving from high-dimensional discovery to tailored, low-plex panels is necessary to use fully automated pipelines—essential for the cost-effective throughput required in multi-site trials.
  • Regulatory Readiness: While currently used for “go/no-go” internal decisions, these exploratory endpoints are paving the way for future Companion Diagnostics (CDx)^12,13.
  • Sample Constraints: Modern workflows are increasingly optimized for low-input or clinical-grade FFPE samples, ensuring that clinical utility is maximized even when tissue is scarce.

Conclusion

The spatial biology toolkit for early clinical research is advancing rapidly, establishing itself as the new standard for solid tumor characterization and delivering insights beyond the reach of traditional methods. 

However, generating the data is only half the battle. Unlocking its true value requires the seamless integration of three complex disciplines: advanced imaging, bioinformatics, and deep biological knowledge. Navigating this intersection demands a skillset that goes beyond standard data analysis.

This is where BioLizard’s expertise makes the difference. We don’t just provide analyses; we provide answers. By leveraging our multi-disciplinary team of Lizards, we help you:

• • Navigate the technical hurdles of scaling from pilot experiments to high-throughput assays for discovery and translational research.
  • Bridge the gap between raw spatial images and statistically sound patient stratification by identifying spatially-aware biomarkers.
  • Uncover the complex spatial signatures that drive treatment success or failure.

Ready to see what your data is really telling you?

Let’s discuss how our spatial workflows can support your path to clinical success.

Contact us >

References:

¹https://www.bio.org/clinical-development-success-rates-and-contributing-factors-2011-2020

²https://recist.eortc.org/

³https://www.cell.com/cancer-cell/fulltext/S1535-6108(25)00543-4

⁴https://clinicaltrials.gov/study/NCT02558140

⁵https://aacrjournals.org/mct/article/15/5/946/176188/RG7386-a-Novel-Tetravalent-FAP-DR5-Antibody

⁶https://www.nature.com/articles/s41598-025-93927-0 

⁷https://clinicaltrials.gov/study/NCT02659631

⁸https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2022.845417/full

⁹https://www.nature.com/articles/s41588-025-02351-7

¹⁰https://pmc.ncbi.nlm.nih.gov/articles/PMC11866545/

¹¹https://aacrjournals.org/clincancerres/article/29/18/3813/728903/Trackable-Intratumor-Microdosing-and-Spatial

¹²https://ir.acrivon.com/news-releases/news-release-details/acrivon-therapeutics-announces-fda-has-granted-breakthrough-0

¹³https://diagnostics.roche.com/global/en/products/lab/pd-l1-sp142-assay-ventana-rtd001231.html

Recognized by

CH Office

BioLizard SA
Digital Health Hub – Biopôle
Bâtiment Phenyl Rte de la Corniche 3 1066 Epalinges, Switzerland