Integrative analyses of single-cell transcriptome and regulome using MAESTRO

Abstract

We present Model-based AnalysEs of Transcriptome and RegulOme (MAESTRO), a comprehensive open-source computational workflow ( http://github.com/liulab-dfci/MAESTRO ) for the integrative analyses of single-cell RNA-seq (scRNA-seq) and ATAC-seq (scATAC-seq) data from multiple platforms. MAESTRO provides functions for pre-processing, alignment, quality control, expression and chromatin accessibility quantification, clustering, differential analysis, and annotation. By modeling gene regulatory potential from chromatin accessibilities at the single-cell level, MAESTRO outperforms the existing methods for integrating the cell clusters between scRNA-seq and scATAC-seq. Furthermore, MAESTRO supports automatic cell-type annotation using predefined cell type marker genes and identifies driver regulators from differential scRNA-seq genes and scATAC-seq peaks.

TISCH: a comprehensive web resource enabling interactive single-cell transcriptome visualization of tumor microenvironment

“…We applied a standardized analysis workflow based on MAESTRO v1.1.0 ( 18 ) for processing all the collected datasets, including quality control, batch effect removal, cell clustering, differential expression analysis, cell-type annotation, malignant cell classification and gene set enrichment analysis (GSEA; Figure 2 ). The raw count, TPM or FPKM table was used as input for the standardized workflow.…”

Chromatin accessibility profiling methods

“…However, due to the substantial scale and sparsity of the region by cell count matrix, specialized bioinformatics tools have been developed -mostly for scATAC-seq data -to handle these assay-specific challenges 191,[233][234][235][236][237][238][239][240][241][242] . One major point in which these tools differ is the way they define genomic regions to be used as features, either as peaks from bulk or aggregated single-cell data (chromVar 239 , Cicero 238 , cisTopic 191 , scABC 241 , Scasat 233 , MAESTRO 242 ), peaks from pseudo-bulk samples 56 or fixed-size bins 56 (SnapATAC 243 ). Another difference between the bulk and single cell-based algorithms is what the count features represent, for example, counting reads in peaks (cisTopic 56,191 , scABC 241 , Scasat 233 , MAESTRO 242 ), counting gapped k-mers under peaks or around transposase cut sites (BROCKMAN 234 , chromVAR 239 ), or counting reads overlapping TF motifs in peaks or genome-wide (chromVar 239 , SCRAT 237 ) 244 .…”

“…One major point in which these tools differ is the way they define genomic regions to be used as features, either as peaks from bulk or aggregated single-cell data (chromVar 239 , Cicero 238 , cisTopic 191 , scABC 241 , Scasat 233 , MAESTRO 242 ), peaks from pseudo-bulk samples 56 or fixed-size bins 56 (SnapATAC 243 ). Another difference between the bulk and single cell-based algorithms is what the count features represent, for example, counting reads in peaks (cisTopic 56,191 , scABC 241 , Scasat 233 , MAESTRO 242 ), counting gapped k-mers under peaks or around transposase cut sites (BROCKMAN 234 , chromVAR 239 ), or counting reads overlapping TF motifs in peaks or genome-wide (chromVar 239 , SCRAT 237 ) 244 . ArchR 236 uses an iterative feature definition method; it first defines a feature-by-cell count matrix of the number of reads per feature (in this case, 500-bp genomic bins) across all single cells, which then undergoes an iterative latent semantic indexing reduction to generate the cell clusters and pseudo-bulk samples on which peaks are called.…”

Unbiased integration of single cell multi-omics data

“…Taking the integration of scRNA-seq and scATAC-seq as an example, the matrix can be derived from scATAC-seq profiles by summing reads in gene bodies 17,19,23 . This can also be input from the regulatory potential (RP) model in MAESTRO 16 . In a simpler case where and have matched features, the integration tasks fall into two categories: 1) batch correction for scRNA-seq data across individuals, species, or technologies; 2) integration of scRNA-seq with spatial transcriptome data.…”

“…This is mathematically challenging, however, as there are many possible ways to simultaneously align a large number of cells and features. To address this challenge, existing computational approaches followed two directions: 1) aligning features empirically before aligning cells [16][17][18][19] ; 2) obtaining separate embeddings for each modality, followed by performing unsupervised manifold alignment [20][21][22] . Taking integration of scRNA-seq and singe cell assay for transposase accessible chromatin sequencing (scATAC-seq) as an example, the first category of methods require constructing a "gene activity matrix" from scATAC-seq data by counting DNA reads aligned near and within each gene 23 .…”