Biology subjects

Eiche, M.

Publications and source records attributed to Eiche, M..

2 recordsLinked to original sources

bioRxiv2026

Crowding governs resistance confinement and guides the improvement of adaptive therapy in a 2D melanoma model system

The emergence of treatment resistant mutants is one of the primary contributors to therapy failure in cancer. Evolution-based adaptive therapies aim to exploit the competition between sensitive and resistant cells to prolong therapeutic control. However, in the densely packed tissues of solid tumors, competition is not only determined by cell-intrinsic factors but also a result of local physical interactions between neighboring cells. The successful development of physics-informed adaptive therapies therefore hinges upon the bottom-up integration of tissue mechanics and experimentally grounded models of tumor evolution. Here, we show how population-level treatment response emerges from mechanical interactions between individual cells. Combining genetically tailored therapy-mimicry experiments in spatially structured 2D melanoma colonies with mechanistic, particle-based tissue simulations as digital twins in a real-to-sim-to-real framework, we reveal how treatment-induced changes in mechanical crowding shape a cascade of resistance expansion, spatial competitive escape, and consequent therapy failure. We then link these underlying physical mechanisms to population-level observables in an abstracted mathematical model for reinforcement-learning-based therapy optimization and a distilled treatment-decision-heuristic, which we test experimentally. Together, our results identify mechanical crowding and escape as governing principles of resistance-mediated therapy failure in solid tumours and provide a basis for mechanistically informed, evolution-based treatment strategies.

biophysics↗

bioRxiv2025

GRAPEVINE: A Reinforcement Failing Framework for AI-Guided Discovery of Hidden Spatial Dynamics in Adaptive Tumor Therapy

Artificial intelligence is revolutionizing scientific discovery in medicine, with reinforcement learning (RL) emerging as a promising tool for optimizing therapeutic strategies. Yet applying RL to complex scenarios such as therapy dynamics in solid tumors is constrained by the challenge of constructing training environments that are both computationally efficient and mechanistically interpretable. Here we introduce Reinforcement Failing, an AI-guided, human-in-the-loop discovery framework that shifts the focus from agent policy optimization to the refinement of the training environment itself. By combining multi-fidelity RL with group-relative performance evaluation across agent cohorts, Reinforcement Failing systematically reveals emergent mechanisms that first-principles models overlook. We apply this framework to adaptive therapy in solid tumors, which seeks to delay resistance-mediated treatment failure. In this setting, Reinforcement Failing uncovered a coupling between the mechanically driven collective motion of cells and spatially-heterogeneous proliferation that strongly influences therapy outcomes. Incorporating these emergent physical mechanisms into an augmented training environment improved cross-environment therapeutic performance and exposed potential pitfalls in translation. More broadly, these findings position Reinforcement Failing as a powerful artificial scientific discovery framework, capable of deciphering high-complexity processes at the interface of physics, machine learning, and medicine.

biophysics↗