#38 · Systems biology

Tractable models infer channel kinetics and oscillator phase dynamics from data

The dominant thread is parameter inference for mechanistic dynamical models. A hybrid framework embeds neural ODEs in conductance-based neuron models to learn unknown ion channel gating kinetics from voltage recordings. A Kuramoto-von Mises time series model proposes a non-equilibrium Langevin-based formulation with closed-form maximum likelihood estimation at high sampling rates for coupled oscillator phases. A stochastic spatiotemporal flux-balance model of oligodendrocyte precursor cell calcium transients integrates membrane voltage and calcium-induced calcium release, reproducing spontaneous and evoked responses. What is new is the emphasis on tractable estimation: a closed-form likelihood in one case, and learned sub-models embedded within biophysical structure in another. The items do not share a single scale or system, so the common thread is methodological rather than biological.