Sparse regression methods for nonlinear system identification received significant attention in recent years. These methods are seen as especially promising for scientific machine learning applications, where the goal is to learn an interpretable dynamic model from a time series of experimental observations. Popular methods including SINDy and weak-form sparse identification address the case where the observed process is deterministic, i.e. can be described as a noisy observation of a system of ordinary or partial differential equations. However, many systems of scientific interest, including many biological systems, have significant process noise and are better described with stochastic differential equation models. In this talk we present a new sparse regression-based system identification approach that allows for the simultaneous learning of nonlinear drift and diffusion terms in stochastic differential equations from data, leveraging classic identities from Itô calculus. Analogously to weak-form methods for deterministic systems, our new method involves numerical integration of stochastic signals against user-defined test functions. We illustrate the effectiveness of our approach on benchmark examples and discuss open questions.