multiSLICE estimates multilayer networks from multimodal data via a shared latent space, with rigorous theoretical guarantees, extensive simulations, a neuroimaging application, and benchmarks against state-of-the-art methods.