The final product of spatial biology is not a map.
It is a decision that others can reproduce and trust.
If the same tissue is analyzed using two spatial transcriptomics platforms, will the results be identical? If the same raw data are processed through different analysis pipelines, will they yield the same cell types and biomarkers?
In practice, results can vary depending on the platform’s measurement principles, gene panel, tissue preprocessing, cell boundary definition, analysis parameters, and algorithms. This does not necessarily mean that one result is wrong. Rather, each technology measures different aspects of biological reality with different emphases.
At the research stage, these differences can themselves be a starting point for new discoveries. But the situation changes when selecting a target or candidate compound in which tens of billions of won will be invested. We must be able to explain why the results differ and demonstrate the conditions under which the same conclusion can be reached.
The final bottleneck to making spatial transcriptomics a standard infrastructure for drug development is not accuracy alone. It is trust in reproducibility, comparability, provenance, and economic viability.
Research Benchmarks and Decision-Making Benchmarks Are Different
For research benchmarks, resolution, the number of detected genes, sensitivity, and cell segmentation accuracy are important. These metrics are essential for understanding a technology’s performance.
Industrial decision-making, however, raises additional questions. Does the prioritization of targets remain consistent when the platform or batch changes? Are the differences between tumor and normal tissues also confirmed at the protein level? Are spatial biomarkers linked to actual clinical outcomes or drug responses? How long does it take to obtain the results, and how much does each decision cost?
The method that offers the highest technical resolution is not necessarily the most suitable for every question. Cost and throughput may be more important in studies comparing large cohorts, whereas depth and precision may matter more when examining cellular interactions within a limited region.
The history of the technology industry has repeatedly shown that the product with the best performance on a single metric does not necessarily become the market standard. Compatibility with existing workflows, ease of use, cost, supply chains, and the broader ecosystem must all work together. Spatial biology is no exception.
RNA-Based Discoveries Must Be Connected to Other Evidence
Spatial transcriptomics is powerful, but RNA expression alone is not sufficient to establish the suitability of a drug target. Because antibodies and cell therapies generally act on proteins, signals detected at the RNA level must be validated for actual protein expression and localization on the cell membrane.
This is why orthogonal validation using immunohistochemistry (IHC) or spatial proteomics is necessary: it provides independent validation through technologies based on different principles. When possible, retrospective validation using clinical outcomes from past patients should also be conducted to assess the relationship between biomarkers and treatment response.
However, the fact that RNA and protein levels do not always align does not diminish the value of spatial transcriptomics. The discrepancies between the two measurements may themselves provide important clues about translational regulation, protein degradation, or tissue states. The key is not to declare one data type as the truth, but to explain where different forms of evidence agree and where they diverge.
What Must Be Standardized Is the Structure of Evidence, Not the Method
There are also arguments against standardization. Fixing formats too early, while the technology is still evolving rapidly, could slow the adoption of new measurement approaches and analytical methods. Nor is it desirable to force all data into a single file format or a single pipeline.
The goal of standardization should not be for every institution to use the same technology. It should be to establish a minimum common foundation that enables results obtained through different methods to be compared and validated.
Reference: https://spatialdata.scverse.org/en/stable/
In the short term, reference tissues and control samples should be used to measure variation across batches and institutions. Analysis provenance should also be standardized so that the data used, QC criteria, parameters, and software versions are documented.
In the medium term, cross-platform benchmarks that analyze the same tissue using multiple platforms should be accumulated alongside RNA–protein validation data. The focus should be not on whether the numerical results from different technologies match perfectly, but on whether they lead to the same decision.
In the long term, pharmaceutical companies, technology companies, hospitals, and regulatory agencies should jointly design benchmarks for decision-making. These benchmarks should measure not only accuracy but also cost per decision, turnaround time, failure rates, and the reproducibility of results.
The Role Portrai Has Chosen
What Portrai has learned in practice is that connecting multiple types of evidence matters more than declaring the superiority of any particular platform. Target candidates may be discovered through spatial RNA data, but they can move into actual development decisions only when comparisons with normal tissues, protein-level validation, consistency across patients, and the feasibility of drug delivery are also established.
We therefore focus on comparing how different data and analyses answer the same question, rather than trying to make one measurement technology the winner. Platforms will continue to evolve, but the evidence supporting a decision must remain traceable and verifiable.
This approach also has limitations. For rare cancers or new modalities, sufficient clinical outcome data may not be available, making retrospective validation impossible. In such cases, rather than concealing incomplete evidence, we must clearly present the scope of uncertainty and the plan for future validation, then continue to strengthen the evidence.
The Next Five Years
In an optimistic scenario, common references and benchmarks will accumulate, allowing results from different platforms to be used together in the drug development process. More realistically, validation frameworks are likely to become established first in selected cancer types and modalities with high clinical value.
Conversely, if companies and institutions continue to claim performance based on different criteria, decision-makers’ trust may fail to keep pace with the growth in data. Mass adoption will begin not with the technology that creates the most sophisticated maps, but with a system that produces decisions people can trust repeatedly.
The final product of spatial biology is not a map. It is a decision that others can reproduce and trust.
In the next article, we will explore three scenarios for 2030 and ask whether spatial transcriptomics can move beyond a distinct field of research to become foundational infrastructure for drug development and precision medicine once the six bottlenecks have been resolved.
