MultiMAP: Dimensionality Reduction and Integration of Multimodal Data

Abstract

Multimodal data is rapidly growing in many fields of science and engineering, including single-cell biology. We introduce MultiMAP, an approach for the dimensionality reduction and integration of multiple datasets. MultiMAP recovers a single manifold on which all of the data resides and then projects the data into a single low-dimensional space so as to preserve the structure of the manifold. MultiMAP is based on a framework of Riemannian geometry and algebraic topology, and generalizes the popular UMAP algorithm to the multimodal setting. MultiMAP can be used for visualization of multimodal data, and as an integration approach that enables joint analyses. MultiMAP has several advantages over existing integration strategies for single-cell data, including that MultiMAP can integrate any number of datasets, leverages features that are not present in all datasets (i.e. datasets can be of different dimensionalities), is not restricted to a linear mapping, can control the influence of each dataset on the embedding, and is extremely scalable to large datasets. We apply MultiMAP to the integration of a variety of single-cell transcriptomics, chromatin accessibility, methylation, and spatial data, and show that it outperforms current approaches in preservation of high-dimensional structure, alignment of datasets, visual separation of clusters, transfer learning, and runtime. On a newly generated single-cell Assay for Transposase-Accessible Chromatin using sequencing (scATAC-seq) and single-cell RNA-seq (scRNA-seq) dataset of the human thymus, we use MultiMAP to integrate cells along a temporal trajectory. This enables the quantitative comparison of transcription factor expression and binding site accessibility over the course of T cell differentiation, revealing patterns of transcription factor kinetics.

UINMF performs mosaic integration of single-cell multi-omic datasets using nonnegative matrix factorization

“…2a). Single-nucleus ATAC data is extremely sparse, with only 1-10% of peaks detected per cell, compared to the 10-45% of genes captured per cell in scRNA sequencing methods 28 . Including the intergenic snATAC data allows more of the detected regions to be used from each cell, a distinct advantage in such sparse datasets.…”

MOJITOO: a fast and universal method for integration of multimodal single cell data

“…We use three distinct metrics to measure the accuracy of the methods. The structure score measures the similarity between two latent space structures 20 . It is based on the Pearson correlation of the pairwise Euclidean…”

“…First, we evaluate algorithms regarding their structure preservation, i.e. the average similarity between the euclidean distances in the shared space and distances in the space of each modality 20 . Results indicate highest structure scores for MOJITOO (4 out of 6) followed by MOFA (2 out of 6).…”

“…The latent spaces are then evaluated with measures regarding the accuracy of clustering (adjusted Rand index), distance (silhouette score) and structure preservation, i.e. relation between shared space and original space of individual modalities 20 . Altogether, results show a superior performance of MOJITOO in both computational requirements and accuracy of estimated latent spaces.…”

sciCAN: Single-cell chromatin accessibility and gene expression data integration via Cycle-consistent Adversarial Network

“…The default architecture of sciCAN shown in Fig. 1 has RNA-seq data playing the central role, primarily because RNA-seq data usually shows greater discriminative power than ATAC-seq in terms of cell-type identification [22][23][24][25][26] . Next, we examined if this setup is critical to good integration by sciCAN.…”