An arXiv preprint gives a formal designability guarantee for a narrow class of RNA secondary structures, while stressing that the result is confined to a simplified pairing model.
The method leaves hard reaction choices intact in the forward simulation, then uses normalized propensities to estimate gradients; its strongest reported speed advantage came in one genetic-oscillator benchmark.
An arXiv preprint presents PEtab SciML, an interoperable format for specifying, sharing and training scientific machine-learning models that combine ordinary differential equations with machine learning.