MarcoPolo: a method to discover differentially expressed genes in single-cell RNA-seq data without depending on prior clustering

Abstract

The standard analysis pipeline for single-cell RNA-seq data consists of sequential steps initiated by clustering the cells. An innate limitation of this pipeline is that an imperfect clustering result can irreversibly affect the succeeding steps. For example, there can be cell types not well distinguished by clustering because they largely share the global structure, such as the anterior primitive streak and mid primitive streak cells. If one searches differentially expressed genes (DEGs) solely based on clustering, marker genes for distinguishing these types will be missed. Moreover, clustering depends on many parameters and can often be subjective to manual decisions. To overcome these limitations, we propose MarcoPolo, a method that identifies informative DEGs independently of prior clustering. MarcoPolo sorts out genes by evaluating if the distributions are bimodal, if similar expression patterns are observed in other genes, and if the expressing cells are proximal in a low-dimensional space. Using real datasets with FACS-purified cell labels, we demonstrate that MarcoPolo recovers marker genes better than competing methods. Notably, MarcoPolo finds key genes that can distinguish cell types that are not distinguishable by the standard clustering. MarcoPolo is built in a convenient software package that provides analysis results in an HTML file.

scMEB: A fast and clustering-independentmethod for detecting differentially expressedgenes in single-cell RNA-seq data

“…Therefore, we can draw similar conclusions about marker gene identification for these two methods of defining marker genes. We also found that MarcoPolo performed differently compared to the results in [14], despite conducting the same study using the same dataset. It is possible that MarcoPolo only used the genes with average expression levels in the top 30th percentile for data preprocessing, which may have led to some DEGs being ignored.…”

“…All three methods were developed for data for which cell clusters were not obvious to detect DEGs without prior clustering. Compared with other methods for detecting DEGs under two conditions, these three methods are faster and more accurate for DEG detection [13,14]. scMEB needs some non-DEGs to build the model; however, the ground truth of the genes is unknown.…”

“…MarcoPolo can involve fitting a mixture model for each gene, which means it takes a long time to estimate the parameters. Because MarcoPolo (GPU version) is 60 ∼ 90 times faster than MarcoPolo (CPU version) [14], it is reasonable to infer that MarcoPolo (GPU version) would finish processing these 11 real datasets in minutes. Haystack and Haystack.adv, both using 50 PCs as inputs, had similar runtimes.…”

“…Cell markers are widely used for labeling cell populations and exploring cell compositions [20]. Kim et al [14] provided two methods to obtain marker genes and compared the performance of marker gene identification through the receiver operating characteristic (ROC) curve and area under the curve (AUC). For the first method, we selected marker genes via the log fold change values.…”

“…Considering the drawbacks, some methods have been proposed to identify DEGs without reference clustering results. Representative methods include singleCellHaystack [13] and MarcoPolo [14]. singleCellHaystack employs Kullback-Leibler divergence to find genes with expressions that are non-uniformly distributed in a low-dimensional space.…”