Back

Interpretable spatial ecological dynamics reveal ecological retention escape and amplification in pediatric leukemia

Kim, S.-H.

2026-07-30 cancer biology
10.64898/2026.07.27.741112 bioRxiv
Show abstract

Spatial transcriptomics enables tumor ecosystems to be examined within their native tissue architecture, yet many computational approaches remain primarily descriptive and provide limited insight into how spatial organization may constrain or facilitate disease persistence. Here, we develop a therapy-aware spatial ecological framework that integrates ecological-context discovery, ecological opportunity landscapes, Ornstein-Uhlenbeck (OU)-like retention, Levy-like escape, and branching-like amplification to characterize pediatric leukemia tissues. Using normal bone marrow reference programs and pediatric leukemia spatial transcriptomic sections, we identify five recurrent candidate ecological contexts with distinct malignant, immune, stromal, vascular, inflammatory, and stress-associated features. These contexts form spatially heterogeneous ecological opportunity landscapes across bone marrow and extramedullary samples. OU-like summaries quantify context-specific attractor position and ecological retention, whereas Levy-like analyses identify rare cross-context displacements consistent with discontinuous ecological remodeling. Branching-like amplification integrates ecological abundance, latent displacement, opportunity, escape, and escaped-state fraction into an interpretable index of ecological expansion. An independent treatment-associated cohort shows that a primitive-state retention proxy, escape, and amplification generate coherent ecological profiles qualitatively consistent with residual persistence and relapse-associated expansion. Because the available spatial datasets are cross-sectional, these quantities are interpreted as operational summaries of tissue organization rather than direct estimates of temporal evolutionary processes. Together, the framework moves spatial transcriptomic analysis beyond static tissue mapping by providing an interpretable representation of ecological constraint, rare displacement, amplification, and therapy-associated persistence in pediatric leukemia.

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.