clusterExperiment and RSEC: A Bioconductor package and framework for clustering of single-cell and other large gene expression datasets

Abstract

Clustering of genes and/or samples is a common task in gene expression analysis. The goals in clustering can vary, but an important scenario is that of finding biologically meaningful subtypes within the samples. This is an application that is particularly appropriate when there are large numbers of samples, as in many human disease studies. With the increasing popularity of single-cell transcriptome sequencing (RNA-Seq), many more controlled experiments on model organisms are similarly creating large gene expression datasets with the goal of detecting previously unknown heterogeneity within cells. It is common in the detection of novel subtypes to run many clustering algorithms, as well as rely on subsampling and ensemble methods to improve robustness. We introduce a Bioconductor R package, , that implements a general and flexible strategy we entitle Resampling-based Sequential Ensemble Clustering (RSEC). RSEC enables the user to easily create multiple, competing clusterings of the data based on different techniques and associated tuning parameters, including easy integration of resampling and sequential clustering, and then provides methods for consolidating the multiple clusterings into a final consensus clustering. The package is modular and allows the user to separately apply the individual components of the RSEC procedure, i.e., apply multiple clustering algorithms, create a consensus clustering or choose tuning parameters, and merge clusters. Additionally, provides a variety of visualization tools for the clustering process, as well as methods for the identification of possible cluster signatures or biomarkers. The R package is publicly available through the Bioconductor Project, with a detailed manual (vignette) as well as well documented help pages for each function.

Trajectory-based differential expression analysis for single-cell sequencing data

“…Note that contrasting the start and endpoints of a lineage is a special case of a more general capability of tradeSeq to compare the mean expression between any two regions of a given lineage. As such, this test can be considered a generalization of cluster-based discrete DE within a lineage (e.g., Risso et al 25 ).…”

“…Specifically, for each gene, we extract a number of fitted values for each lineage (100 by default). We can then use resampling-based sequential ensemble clustering (RSEC), as implemented in clusterExperiment 25 , to perform the clustering based on (the top principal components of) the standardized fitted values matrix (i.e., the fitted values are standardized to have zero mean and unit variance across cells for each gene). Importantly, we allow for any clustering algorithm that is built-in into cluster-Experiment or chosen by the user to perform the clustering.…”

“…Dimensionality reduction is performed through PCA on the normalized logtransformed counts that are offset by 1 to avoid taking the log of zero, i.e., log ðy þ 1Þ. Clustering is performed through k-means on the first 50 principal components by varying the number of clusters k ∈ {4, …, 15}; stable clusters are derived using clusterExperiment 25 , yielding a final repertoire of 13 cell clusters. Next, slingshot is used to infer trajectories with the initial cluster chosen by known marker genes of horizontal basal cells (HBC), an adult stem cell population.…”

mbkmeans: fast clustering for single cell data using mini-batchk-means

“…The algorithm partitions N cells into k clusters each represented by a centroid, or mean profile, for the cells in the k th cluster. This algorithm is commonly used not only on its own, but also as a component of ensemble clustering [9,10].…”

“…For large enough data, k -means can be slow or completely fail if a user lacks sufficient computational resources. Ensemble clustering approaches that depend on the use of k -means [9,10] run it multiple times (e.g., with different parameter values or on a different data subset) limiting the usability of these packages for large scRNA-seq datasets [11]. We note that our goal here is not to debate the relative merits of k -means as a clustering algorithm -k -means is a well-established method, which has been thoroughly investigated [12] -but to provide users with the ability to use the popular k -means algorithm on large single-cell datasets.…”