Iterative Refinement of Cellular Identity from Single-Cell Data Using Online Learning

Recent experimental advances have enabled high-throughput single-cell measurement of gene expression, chromatin accessibility and DNA methylation. We previously used integrative non-negative matrix factorization (iNMF) to jointly learn interpretable low-dimensional representations from multiple single-cell datasets using dataset-specific and shared metagene factors. These factors provide a principled, quantitative definition of cellular identity and how it varies across biological contexts. However, datasets exceeding 1 million cells are now widely available, creating computational barriers to scientific discovery. For instance, it is no longer feasible to analyze large datasets using standard pipelines on a personal computer with limited memory capacity. Moreover, there is a need for an algorithm capable of iteratively refining the definition of cellular identity as efforts to create a comprehensive human cell atlas continually sequence new cells. To address these challenges, we developed an online learning algorithm for integrating large and continually arriving single-cell datasets. We extended previous online learning approaches for NMF to minimize the expected cost of a surrogate function for the iNMF objective. We also derived a novel hierarchical alternating least squares algorithm for iNMF and incorporated it into an efficient online algorithm. Our online approach accesses the training data as mini-batches, decoupling memory usage from dataset size and allowing on-the-fly incorporation of new datasets as they are generated. The online implementation of iNMF converges much more quickly using a fraction of the memory required for the batch implementation, without sacrificing solution quality. Our new approach processes 1.3 million single cells from the entire mouse embryo on a laptop in 25 minutes using less than 500 MB of RAM. We also analyze large datasets without downloading them to disk by streaming them over the internet on demand. Furthermore, we construct a single-cell multi-omic cell atlas of the mouse motor cortex by iteratively incorporating eight single-cell RNA-seq, single-nucleus RNA-seq, single-nucleus ATAC-seq, and single-nucleus DNA methylation datasets generated by the BRAIN Initiative Cell Census Network. Our approach obviates the need to recompute results each time additional cells are sequenced, dramatically increases convergence speed, and allows processing of datasets too large to fit in memory or on disk. Most importantly, it facilitates continual refinement of cell identity as new single-cell datasets from different biological contexts and data modalities are generated.