Network embedding approaches are gaining momentum to analyse a large variety\nof networks. Indeed, these approaches have demonstrated their efficiency for\ntasks such as community detection, node classification, and link prediction.\nHowever, very few network embedding methods have been specifically designed to\nhandle multiplex networks, i.e. networks composed of different layers sharing\nthe same set of nodes but having different types of edges. Moreover, to our\nknowledge, existing approaches cannot embed multiple nodes from\nmultiplex-heterogeneous networks, i.e. networks composed of several layers\ncontaining both different types of nodes and edges. In this study, we propose\nMultiVERSE, an extension of the VERSE method with Random Walks with Restart on\nMultiplex (RWR-M) and Multiplex-Heterogeneous (RWR-MH) networks. MultiVERSE is\na fast and scalable method to learn node embeddings from multiplex and\nmultiplex-heterogeneous networks. We evaluate MultiVERSE on several biological\nand social networks and demonstrate its efficiency. MultiVERSE indeed\noutperforms most of the other methods in the tasks of link prediction and\nnetwork reconstruction for multiplex network embedding, and is also efficient\nin the task of link prediction for multiplex-heterogeneous network embedding.\nFinally, we apply MultiVERSE to study rare disease-gene associations using link\nprediction and clustering. MultiVERSE is freely available on github at\nhttps://github.com/Lpiol/MultiVERSE.\n