Ask a question.
Get a spatial analysis.
Node turns uploaded spatial data into cell-level context and, with a matching high-resolution H&E image, automatically annotates cells and tissue structures. Ask a research question; the agent uses that context to assemble an analysis from Node's functions. Checks and your approval come before execution; the results return with interpretation.
The H&E image becomes part of the analysis.
When a high-resolution H&E image of the assayed section is included, Node automatically annotates cells and tissue structures. Those annotations are available to the agent as features for analysis—not just an overlay in the viewer.
Same section. High resolution.
The H&E image and spatial assay refer to the same tissue section, so morphology can inform questions about its molecular data.
Cells and regions identified.
AI-based pathology creates cell and tissue-region annotations that you can inspect in the sample viewer.
Use the annotations.
The agent can use those labels in question-specific analyses, including differential gene expression across annotated regions.
Six markers. One tissue.
This example maps six genes back to their positions in the tissue and reports the percentage of cells expressing each marker. It shows molecular output; H&E-derived annotations are not shown in this figure.
- CDH116.3%
- COL1A132.5%
- COL1A227.8%
- COL5A117.4%
- COL5A29.7%
- THY110.4%
Percentages are from the example figure. Bar lengths are relative to the largest value shown, 32.5%.
From sample context to interpretation.
Sample preparation happens automatically. The researcher enters with a question, reviews the proposed plan, and decides what runs.
The sample becomes context
Behind the scenes, Node builds cell-level expression, cell-type, and spatial profiles. With a matching high-resolution H&E image, it also annotates cells and tissue structures.
The question shapes the analysis
Ask about a gene, cell type, or spatial pattern. The agent combines that question with the prepared sample and assembles a custom analysis from Node's defined Python functions.
The plan is checked and approved
Compatibility checks send invalid module calls back for revision. You can inspect the steps, adjust settings such as thresholds, and approve what runs.
The results return with interpretation
The report includes plots, data, and methods. The agent uses the results and available sample context to generate an interpretation and suggested next steps.
A custom analysis, not an improvised script.
Node's library contains dozens of high-level Python functions with defined interfaces. The agent selects and connects them using the sample context and your question. When calls are incompatible, the system sends them back for revision before you review the plan.
Results with their workings.
The report is meant to be examined, not just skimmed for a conclusion. It keeps the original question, analysis plan, methods, outputs, and interpretation together.
See the plots and data.
Review spatial figures from completed steps and download the underlying outputs for further work.
Follow how they were made.
The original question and approved plan sit alongside the methods used to generate the results.
Decide what to test next.
An AI-generated interpretation summarizes the results and suggests next steps for the research team to evaluate.
See the platform in motion.
YouTube does not allow its player to run reliably inside a downloaded HTML file. The video opens on YouTube.