SCIM: universal single-cell matching with unpaired feature sets

Abstract

Motivation Recent technological advances have led to an increase in the production and availability of single-cell data. The ability to integrate a set of multi-technology measurements would allow the identification of biologically or clinically meaningful observations through the unification of the perspectives afforded by each technology. In most cases, however, profiling technologies consume the used cells and thus pairwise correspondences between datasets are lost. Due to the sheer size single-cell datasets can acquire, scalable algorithms that are able to universally match single-cell measurements carried out in one cell to its corresponding sibling in another technology are needed. Results We propose Single-Cell data Integration via Matching (SCIM), a scalable approach to recover such correspondences in two or more technologies. SCIM assumes that cells share a common (low-dimensional) underlying structure and that the underlying cell distribution is approximately constant across technologies. It constructs a technology-invariant latent space using an autoencoder framework with an adversarial objective. Multi-modal datasets are integrated by pairing cells across technologies using a bipartite matching scheme that operates on the low-dimensional latent representations. We evaluate SCIM on a simulated cellular branching process and show that the cell-to-cell matches derived by SCIM reflect the same pseudotime on the simulated dataset. Moreover, we apply our method to two real-world scenarios, a melanoma tumor sample and a human bone marrow sample, where we pair cells from a scRNA dataset to their sibling cells in a CyTOF dataset achieving 90% and 78% cell-matching accuracy for each one of the samples, respectively. Availability and implementation https://github.com/ratschlab/scim. Supplementary information Supplementary data are available at Bioinformatics online.

Machine learning for single cell genomics data analysis

“…Several ML approaches have been developed for that purpose, for instance by characterizing cells across measurements, projecting multiple measurements into a common latent space or learning the missing modalities. Transcriptomics is typically one of the modalities that is integrated, together with chromatin accessibility [69, 52, 70], DNA [71], DNA methylation [72, 52], proteomic data [73, 74, 69, 75, 76] or CRISPR perturbations [77 • , 78].…”

“…While these methods can in theory be applied to any bi-modal omics dataset, hyperparameter selection is difficult when no co-assay data is available for MMD-MA, SCOT and UnionCOM. Among models that do not require co-assay data, some use weak supervision such as SCIM [72], an adversarial AE model that assumes that the cell types are known for a fraction of the cells and Seurat v3 [52], a canonical correlation analysis (CCA)-based model that relies on building anchor cells using mutual nearest neighbours. Applied to single-cell CRISPR screenings, scMAGeCK [78] relies on statistical analyses and MUSIC [77 • ] on topic modeling in order to link gene perturbations to cell phenotype.…”

Multi-omics integration and regulatory inference for unpaired single-cell data with a graph-linked unified embedding framework

“…At the omics feature level, presumed feature interactions have been used via feature conversion 15,16 or coupled matrix factorization 18,19 . At the cell level, known cell types have also been used via (semi)supervised learning 46,47 , but this approach incurs substantial limitations in terms of applicability since such supervision is typically unavailable and in many cases serves as the purpose of multi-omics integration per se 25 . Notably, one of these methods was proposed with a similar autoencoder architecture and adversarial alignment 47 , but it relied on matched cell types or clusters to orient the alignment.…”

Deep Learning Applications in Single-Cell Omics Data Analysis

“…Single-Cell Data Integration via Matching (SCIM) (Stark, Ficek et al 2020) is another deep generative approach that constructs a technology-invariant latent space to recover cell correspondences among datasets, even with unpaired feature sets. The architecture is a modified auto-encoder with an integrated discriminator network, similar to the one in GANs, allowing the network to be trained in an adversarial manner.…”