From Organoid to Original Tumor: How iSCanGuide™ Put Its Own Model to the Test

By Cristiana Iosef, D.V.M/Ph.D., Principal Life Scientist

A glioblastoma (GBM) organoid grown in a dish is a bet. The bet is that a few millimeters of lab-grown tissue actually behaves like the tumor it is meant to stand in for, that its architecture, its signaling, its invasive edge all mirror what is happening inside a patient. Most of the time that bet is taken on faith.

Here’s the finding up front: in a new re-analysis of data from the GLORIA glioblastoma clinical trial (Phase I/II, NCT04121455), Advaita’s team found that a GBM organoid’s spatial architecture and its NOTCH1–DLL3/4 signaling matched six independent primary GBM tumors. The organoid behaved like the disease it’s meant to represent, and the pictures make the case.

Figure 1. Concentric architecture in a GBM organoid. The same core/rim landmarks were used to annotate spatial transcriptomics across the study. The top picture shows how cell populations cluster concentrically. The bottom picture left NOTCH1 positive cells and gene transcripts density, the bottom picture right shows the delineation between the core of the organoid and its rim, with a display of key markers transcript densities.

Using iSCanGuide™, Advaita’s spatial and single-cell analysis platform, together with iPathwayGuide™, the team ran the GLORIA trial’s Xenium spatial transcriptomics data back through the pipeline alongside organoids grown in Core2Edge™ system, GBM organoids cultured directly on slices of healthy brain tissue to recreate the tumor–host interface. The question was simple: does the organoid’s spatial architecture match the primary tumor?

An organoid organizes itself like a tumor. The first thing iSCanGuide™’s spatial mapping showed was structure. GBM organoids in the Core2Edge system don’t grow as an undifferentiated blob; instead they self-organize into concentric zones, a dense core wrapped in an active, invasive rim.

Zoom into that rim, and you can watch individual cells breaking away from it and migrating into the surrounding healthy tissue which could be the organoid equivalent of invasion.

Figure 2. SOX2 (embryonic marker) expression highlighting a stem-like, embryonic signature in cells caught mid-invasion into the healthy tissue.

Cell–cell communication analysis in iSCanGuide™ ranked the ligand–receptor pairs most active at that invasive boundary, and one signal rose to the top in the organoid: NOTCH1–DLL3/4, a spatially coherent cancer stem cell “escape” program (Moran’s I = 0.814, a strong, non-random spatial pattern). iPathwayGuide™ returned the same program independently.

That alone would be an intriguing organoid finding. What made it a fidelity result is what happened next: the team tested the organoid-derived marker signature against six independent primary GBM tumors from a different patient cohort. The NOTCH1–DLL3/4 program wasn’t just present but it mapped to the same tumor-rim compartment, consistently, across every patient.

By the numbers

• Moran’s I = 0.814 for the NOTCH1–DLL3/4 rim signal (strong, non-random spatial clustering)
• 89 markers in the organoid-derived rim-invasiveness signature
• 6 independent primary GBM tumors tested, from a different patient cohort
• 16M+ cells analyzed across those six tumors

Public data sets: GSM9809293, -4, -5, -6, -7, -8

Figure 3. In primary patient tumors, NOTCH1^DLL4/3 signaling concentrates at the invasive rim while VEGF^CD44 (a component of the CXCL12-(VEGF^CD44)/CXCR4 signaling loop) dominates the tumor core. These are two distinct, spatially separated escape programs, detected in the organoid but confirmed independently of the organoid data.

A second, complementary program, CXCL12-(VEGF^CD44)/CXCR4, emerged from the tumor core, pointing to a distinct vasculogenic progenitor niche dominated by VEGF signaling. Between the two, the primary tumor appears to run two separate escape routes at once: a core process built on CXCL12-(VEGF^CD44)/CXCR4 and VEGF signaling, and a rim process built on NOTCH1^DLL3/4. The GLORIA trial’s anti-angiogenic backbone already targets the first; pairing it with NOTCH1^DLL3/4 blockade could give a rationale for hitting the second as well, a combination therapy addressing both tumor compartments at once.

Figure 4. DLL4, NOTCH1, SOX2, and LAMA2 expression side by side in a primary lesion versus lineage nodes identified by time inference.

Perhaps the most striking image in the dataset came from comparing the primary lesion (the organoid itself) to lineage-disseminated cells, which show first the interaction between the tumoral organoid and the healthy brain tissue and may model a possible relapse after treatment.

The lineage-derived cellular nodes re-express the same DLL4/NOTCH1/SOX2 program seen at the primary rim, which could be a spatial echo of cancer cell versus healthy microenvironment, captured directly in tissue rather than inferred after the fact.

iSCanGuide™ held an organoid dataset and a set of independent primary-tumor datasets to the same analytical standard: same clustering, same cell–cell communication scoring, same pathway interpretation at a scale spanning tens of thousands of organoid cells and well over a million patient cells, and the answers agreed. That’s the real test of the tool, not just the tumor. The organoid’s rim signature wasn’t a lab artifact; it was biology the primary tumors were already showing, independently confirmed by iPathwayGuide™.

That’s the real result here: A GPU-accelerated spatial platform (iSCanGuide™) that seamlessly bridges benchtop organoids and multi-cohort clinical datasets, identifying the same disease process across both settings with sufficient resolution to prioritize whether a combination therapy should target the core, the rim, or both.Do you want this kind of validation on your own data? Book a demo of iSCanGuide.

What is Core2Edge?

Core2Edge™ is a novel organoid culture system. It grows GBM organoids directly on slices of healthy brain tissue to recreate the tumor–host interface, so the invasive edge of the tumor can be studied in a spatial context similar to a real brain.

What did the GLORIA trial re-analysis find?

Re-analyzing Xenium spatial transcriptomics data from the GLORIA glioblastoma trial (Phase I/II, NCT04121455) alongside Core2Edge organoid data, Advaita found that a GBM organoid’s spatial architecture and NOTCH1–DLL3/4 signaling matched six independent primary GBM tumors from a separate patient cohort, evidence that the organoid model reproduces real tumor biology rather than lab-specific artifacts.

Why does the core-versus-rim distinction matter for treatment?

The tumor core runs a CXCL12–(VEGF^CD44)/CXCR4 program tied to VEGF-driven vasculogenesis, while the invasive rim runs a separate NOTCH1^DLL3/4 program. The GLORIA trial’s anti-angiogenic backbone (bevacizumab) already addresses the core process; the rim process is currently untargeted, suggesting combination therapy could hit both tumor escape routes at once.

What’s the difference between iSCanGuide™ and iPathwayGuide™?

iSCanGuide™ is Advaita’s spatial and single-cell analysis platform, used here to map cell clusters, spatial architecture, and cell–cell communication. iPathwayGuide™ is Advaita’s pathway analysis tool, used to independently confirm the same NOTCH1–DLL3/4 signaling program from a pathway-level perspective.

Analysis conducted using iSCanGuide and iPathwayGuide on GLORIA trial Xenium public datasets (GSM9572842 / GSM9572843) and Core2Edge™ organoid data, benchmarked against six independent primary GBM tumor datasets (GSM9809293, -4, -5, -6, -7, -8). Spatial findings are hypothesis-generating; experimental validation is ongoing.

Reference: Giordano FA, Layer JP, Turiello R, et al. L-RNA aptamer-based CXCL12 inhibition combined with radiotherapy and bevacizumab in newly-diagnosed glioblastoma: expansion of the phase I/II GLORIA trial. Nat Commun. 2026;17:3405. https://doi.org/10.1038/s41467-026-71362-7

.