“…Here, integration methods aim to merge individual layers of single-cell data (e.g. transcriptomic, proteomic, epigenomic) * Correspondence to Natalie Stanley, email: natalies@cs.unc.edu and Jeremy Purvis, email: purvisj@email.unc.edu into a unified consensus representation, such as an integrated graph [25] or a joint-embedding [24,26]. To do so, computational approaches have leveraged techniques, including kernel learning [27,28], matrix factorization [29,30,31,32,33], or deep learning [34].…”