This study assesses the feasibility of latent factor analysis via dynamical systems (LFADS) for evaluating differences in the observed spiking response dynamics imposed by two electrical microstimulation regimes in awake rats. LFADS is a recently-developed deep learning method that uses stimulus-aligned neural spiking data to determine the initial neural state of each trial, as well as infer a set of time-dependent perturbations to the learned neural dynamics within trials. We show that time-dependent perturbations inferred by an LFADS model trained on spikes from trials on a single session can distinguish between different stimulation conditions. Furthermore, we use these data to exemplify how LFADS inferences track the evolution of stimulus-related spiking responses during chronic microstimulation experiments.