Solid Tumour Heterogeneity

Solid tumour heterogeneity creates local treatment failure

Solid tumours are spatially complex. Within the same tumour, one region may express a therapy target strongly, while another region may express it weakly or not at all. Some regions may be accessible to engineered T cells, while deeper or more resistant regions may remain difficult to reach. Other regions may be recognised by the therapy but still survive because their tumour state makes them harder to kill.

This creates a key development challenge for solid tumour T-cell therapy: treatment response is shaped by local tumour heterogeneity, not only by the average expression of a selected target.

This turns spatial heterogeneity into a set of practical development questions:

Is the selected target spatially broad enough?
Would a multi-antigen design reduce local escape?
Are residual tumour regions more likely driven by poor target coverage, limited penetration, or resistant tumour states?
Which mechanism should be tested next in experimental models?
Which tumour profiles may be most suitable for a given therapy concept?

Apple Twin is building spatial simulation tools around these questions: mapping where a therapy concept may be strong, where escape may occur, and which mechanism may be responsible. This creates a more practical foundation for discussing target choice, multi-antigen design, tumour penetration, combination strategies, and partner-led validation.

Mapping the key layers of solid tumour response

Apple Twin’s framework is being developed to explore how heterogeneous tumour regions behave over time under different treatment concepts or biological conditions. The model is designed to connect spatial tumour structure with therapy-relevant mechanisms, so that treatment response can be interpreted locally rather than only as a single average outcome.

The framework explores six connected layers:

Tumour heterogeneity
How local tumour regions differ in abundance, growth behaviour, tumour state, and resistance-related features.

Antigen coverage
Whether a selected target, or combination of targets, covers the right tumour regions spatially.

Treatment recognition logic
How different therapy concepts may behave, including single-target, multi-target, logic-gated, or pooled T-cell strategies.

Access and penetration
Whether therapy effect may be limited by local accessibility, tumour-region position, or reduced exposure in harder-to-reach areas.

Local response over time
How tumour regions may shrink, persist, or expand as the simulation evolves under a selected treatment or condition.

Residual disease and escape mechanisms
Whether remaining tumour regions are more likely associated with poor target coverage, limited access, resistant tumour state, or mixed mechanisms.

Together, these layers help turn spatial heterogeneity into clearer therapy-design questions: what is being targeted, what is being reached, what remains, and what mechanism may explain the difference.

Developing a spatial simulation framework

Apple Twin is developing a spatial simulation framework for modelling heterogeneous solid tumour behaviour over time. Starting from public high-grade serous ovarian cancer spatial transcriptomics data, we are building a non-confidential demonstration environment that represents tumour tissue as a spatial map of local biological regions.

Each region can be assigned features such as tumour abundance, antigen-expression pattern, tumour-state activity, accessibility, and resistance-related signals. The model can then simulate tumour development over time under different biological assumptions and therapy conditions.

This allows us to explore how specific treatment concepts may interact with tumour heterogeneity. For example, the framework can compare single-antigen targeting, multi-antigen recognition, logic-gated designs, pooled T-cell strategies, or conditions that alter tumour growth, local killing sensitivity, and treatment access.

The goal is to produce interpretable spatial outputs: where tumour regions may respond, where residual disease may persist, and what local mechanism may explain treatment escape. This provides a practical foundation for collaborative case-study development with partners working on real targets, constructs, tumour models, assay data, and translational decisions.

From simulation comparison to mechanistic insight

Comparing settings, treatments, and biological conditions

The public-data demonstration can show how the same heterogeneous tumour map behaves under different simulated settings. Starting from one baseline spatial tumour map, the framework can apply different treatment concepts or biological conditions, then compare how tumour regions change over time. This could include untreated tumour development, single-target therapy, multi-target recognition, logic-gated designs, pooled therapy concepts, or altered assumptions around tumour growth, access, penetration, and treatment sensitivity.

This comparison is useful because each setting is tested against the same starting tumour landscape. That makes it possible to ask whether a change in outcome is linked to the treatment concept itself, the spatial antigen pattern, the accessibility of tumour regions, or the resistance state of local tumour areas. The output can be compared at several levels: overall tumour burden, spatial response pattern, residual tumour location, size of escape regions, antigen coverage, access limitation, and tumour-state composition.

The goal is not only to ask which setting performs better, but to understand what changes when the setting changes. For example, if multi-target recognition reduces residual regions that remain after single-target therapy, the result may point to antigen coverage as a limiting factor. If broader target recognition improves antigen coverage but residual tumour still remains, the framework can help ask whether the remaining limitation is tumour access, penetration, resistant tumour state, or a combination of mechanisms.

Interpreting biological mechanisms

The public demonstration can also show how simulation outputs can be interpreted mechanistically. Local tumour response depends on a chain of events: the tumour region must be present, recognised by the therapy, reached by the treatment effect, sensitive enough to be controlled, and not growing faster than the therapy can suppress it. When tumour remains, the framework examines which part of this chain may be weak.

For example, residual tumour with low recognition may suggest poor target coverage or antigen escape. Residual tumour with good recognition but low exposure may suggest access or penetration limitation. Residual tumour with good recognition and exposure but low sensitivity may suggest tumour-state-mediated resistance. A restrictive logic-gated design may miss tumour regions that express only one required antigen, suggesting therapy-logic mismatch. When several weak factors overlap, the result may indicate mixed escape.

This turns the public demonstration into more than a visual simulation. It shows how spatial modelling can connect treatment settings to possible biological explanations: where tumour regions respond, where they persist, and whether the likely bottleneck is target coverage, access, resistance biology, therapy logic, growth behaviour, or a combination of mechanisms.

Advancing the framework into applied pilot studies

Apple Twin is developing this framework toward applied pilot studies focused on concrete therapy-design and tumour-heterogeneity questions. These pilots can be built using public datasets, internal modelling scenarios, collaborator data, or industry research questions, depending on the level of access and collaboration available.

Target coverage analysis
Evaluate whether a selected antigen is spatially broad enough across heterogeneous tumour regions. This can help identify whether a single target may leave antigen-low or antigen-negative tumour compartments that could drive residual disease.

Multi-antigen design comparison
Compare single-target, dual-target, multi-target, logic-gated, or pooled therapy concepts on the same tumour landscape. The aim is to understand whether adding targets improves spatial coverage, whether a logic gate is too restrictive, or whether a pooled strategy may address different tumour compartments.

Penetration and access modelling
Explore whether predicted treatment failure is mainly caused by poor access to certain tumour regions. This can support questions around trafficking, tumour-core penetration, local delivery, stromal barriers, or accessibility-related limitations.

Escape-region analysis
Identify where residual tumour regions remain after simulated treatment and classify possible escape mechanisms. Residual regions can be interpreted in terms of target coverage, access limitation, resistant tumour state, therapy-logic mismatch, growth behaviour, or mixed escape.

2D-to-3D simulation extension
Use the public-data 2D spatial framework as a starting point for deeper 3D modelling. Selected tumour regions or representative escape scenarios could be translated into 3D simulations to explore penetration, tumour depth, spatial contact, and local treatment dynamics more realistically.

Mechanistic AI modules
Develop AI-enhanced modules to improve specific parts of the simulation while keeping the model interpretable. These could include tumour-state learning, antigen-combination ranking, resistance modelling, parameter estimation, uncertainty scoring, or therapy-design optimisation.

Interested in spatial simulation for solid tumour T-cell therapy?

Apple Twin is developing spatial modelling tools to explore how tumour heterogeneity may shape therapy response, residual disease, and local escape. We are open to discussions around pilot studies, research collaboration, and applied modelling for solid tumour cell therapy development.

Get Started

Modelling solid tumour heterogeneity for next-generation T-cell therapy design

Apple Twin is developing a spatial simulation framework to explore how CAR-T and other engineered T-cell therapies may perform across heterogeneous solid tumour landscapes.

Solid tumours are spatially heterogeneous. Different regions of the same tumour may vary in antigen expression, tumour state, accessibility, resistance biology, and local sensitivity to immune-cell killing.

Apple Twin is building a spatial simulation framework that converts this heterogeneity into therapy-design insight. Our current public-data demonstration uses spatial transcriptomics to explore how tumour-region maps, antigen-coverage logic, accessibility features, and resistance-related states can be organised into a structured simulation workflow.

This work provides a foundation for collaborative applied case studies with T-cell therapy developers, translational research teams, and academic partners. Together, we can apply the framework to specific targets, constructs, tumour models, assay data, and development decisions.