Computational Methods for Single-Cell Multi-Omics Integration and Alignment

Abstract

Recently developed technologies to generate single-cell genomic data have made a revolutionary impact in the field of biology. Multi-omics assays offer even greater opportunities to understand cellular states and biological processes. The problem of integrating different omics data with very different dimensionality and statistical properties remains, however, quite challenging. A growing body of computational tools is being developed for this task, leveraging ideas ranging from machine translation to the theory of networks, and represents another frontier on the interface of biology and data science. Our goal in this review is to provide a comprehensive, up-to-date survey of computational techniques for the integration of single-cell multi-omics data, while making the concepts behind each algorithm approachable to a non-expert audience.

Methods and applications for single-cell and spatial multi-omics

“…Several alternative architectures have been proposed that optimize for different criteria such as robustness to dropouts and batch effects, disentanglement of the learned latent factors for improved interpretability and cross-modal translation for missing modality imputation 145 , 146 . For further details we refer the reader to a recent overview of proposed deep learning approaches 147 .…”

Application of Deep Learning on Single-Cell RNA Sequencing Data Analysis: A Review

“…For joint-modality sequencing, to leverage both datasets simultaneously, most methods employ a method of joint representation learning, or finding a shared latent representation of the data for all modalities. One challenge with this type of joint sequencing is that there can be an increase in noise and sparsity in the data, compared to scRNA-seq data using one modality [177] . In addition, it is difficult for the balance of both modalities during the embedding process, and it is possible that some modalities can dominate the downstream embedding tasks leading to the reduction of biological variability that exists within one modality.…”

Paired single-cell multi-omics data integration with Mowgli

“…Integrative Matrix Factorization (integrative MF) and variational autoencoders perform dimensionality reduction, jointly embedding the highdimensional multi-omics cellular profilings into a shared lower-dimensional latent space by leveraging common cells/observations 13,26 . Integrative MF, due to its linear nature, defines a latent space with a natural biological interpretation, but it is too simple to catch complex biological processes 13,26 . On the other hand, non-linear methods, as variational autoencoders, have shown great potential in clustering cells, but despite recent works on the subject 27,28 , they inherently lack biological interpretability.…”

“…State-of-the-art multiview learning methods for single-cell multi-omics integration are based on integrative Matrix Factorization 14,19,22 , k-nearest neighbors 15 , or variational autoencoders [16][17][18] . Integrative Matrix Factorization (integrative MF) and variational autoencoders perform dimensionality reduction, jointly embedding the highdimensional multi-omics cellular profilings into a shared lower-dimensional latent space by leveraging common cells/observations 13,26 . Integrative MF, due to its linear nature, defines a latent space with a natural biological interpretation, but it is too simple to catch complex biological processes 13,26 .…”