Biology subjects

Aif, S.

Publications and source records attributed to Aif, S..

3 recordsLinked to original sources

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↗

bioRxiv2024

Beyond comparison: Brillouin microscopy and AFM-based indentation reveal divergent insights into the mechanical profile of the murine retina

Mechanical tissue properties increasingly serve as pivotal phenotypic characteristics that are subject to change during development or pathological progression. The quantification of such material properties often relies on physical contact between a load-applying probe and an exposed sample surface. For most tissues, these requirements necessitate animal sacrifice, tissue dissection and sectioning. These invasive procedures bear the risk of yielding mechanical properties that do not portray the physiological mechanical state of a tissue within a functioning organism. Brillouin microscopy has emerged as a non-invasive, optical technique that allows to assess mechanical cell and tissue properties with high spatio-temporal resolution. In optically transparent specimens, this technique does not require animal sacrifice, tissue dissection or sectioning. However, the extent to which results obtained from Brillouin microscopy allow to infer conclusions about potential results obtained with a contact-based technique, and vice versa, is unclear. Potential sources for discrepancies include the varying characteristic temporal and spatial scales, the directionality of measurement, environmental factors, and mechanical moduli probed. In this work, we addressed those aspects by quantifying the mechanical properties of acutely dissected murine retinal tissues using Brillouin microscopy and atomic force microscopy (AFM)-based indentation measurements. Our results show a distinct mechanical profile of the retinal layers with respect to the Brillouin frequency shift, the Brillouin linewidth and the apparent Youngs modulus. Contrary to previous reports, our findings do not support a simple correlative relationship between Brillouin frequency shift and apparent Youngs modulus. Additionally, the divergent sensitivity of Brillouin microscopy and AFM-indentation measurements to cross-linking or changes post mortem underscores the dangers of assuming both methods can be generally used interchangeably. In conclusion, our study advocates for viewing Brillouin microscopy and AFM-based indentation measurements as complementary tools, discouraging direct comparisons a priori and suggesting their combined use for a more comprehensive understanding of tissue mechanical properties.

biophysics↗

bioRxiv2022

Evolutionary rescue is promoted in compact cellular populations

Mutation-mediated drug resistance is one of the primary causes for the failure of modern antibiotic or chemotherapeutic treatment. Yet, in the absence of treatment many drug resistance mutations are associated with a fitness cost and therefore subject to purifying selection. While, in principle, resistant subclones can escape purifying selection via subsequent compensatory mutations, current models predict such evolutionary rescue events to be exceedingly unlikely. Here, we show that the probability of evolutionary rescue, and the resulting long-term persistence of drug resistant subclones, is dramatically increased in dense microbial populations via an inflation-selection balance that stabilizes the less-fit intermediate state. Tracking the entire evolutionary trajectory of fluorescence-augmented "synthetic mutations" in expanding yeast colonies, we trace the origin of this balance to the opposing forces of radial population growth and a clone-width-dependent weakening of selection pressures, inherent to crowded populations. Additionally conducting agent-based simulations of tumor growth, we corroborate the fundamental nature of the observed effects and demonstrate the potential impact on drug resistance evolution in cancer. The described phenomena should be considered when predicting the evolutionary dynamics of any sufficiently dense cellular populations, including pathogenic microbial biofilms and solid tumors, and their response to therapeutic interventions. Our experimental approach could be extended to systematically study rates of specific evolutionary trajectories, giving quantitative access to the evolution of complex adaptations.

biophysics↗