#15 · Single-cell
Systematic evaluation of analytical choices in single-cell RNA-seq analysis
The dominant thread is systematic evaluation of analytical choice in scRNA-seq workflows, with several items pairing benchmarks with new methods. A comparison of 16 feature selection methods on PBMC3K and Visium data reported a mean overlap of only 23.7% between selected gene sets, with 1,150 genes selected consistently. A systematic evaluation of 47 highly variable gene selection methods across 19 datasets and 18 criteria proposed the hybrid mixHVG approach with an open-source R package. Newer methods target specific sources of variation: COBRA performs linear-model batch correction in full gene space, CellMentor uses cell-type labels in a supervised NMF objective, and RUVIII PBPS controls FDR in pseudobulk analysis without technical replicates. CANTAO introduces average overlap to guide clustering resolution and annotation.
Published papers
CellMentor: cell-type aware dimensionality reduction for single-cell RNA-sequencing data.
CellMentor: a supervised NMF-based dimensionality reduction method that uses cell-type labels in its objective to improve scRNA-seq cell type separation, batch…
CANTAO: guiding clustering and annotation in single-cell RNA sequencing using average overlap.
CANTAO introduces the average overlap metric to compare ranked differentially expressed gene lists for guiding clustering resolution and annotation of…
Removal of unwanted variation in pseudobulk analysis of single-cell RNA sequencing data and the leveraging of pseudoreplicates.
Benchmarks RUV2, RUVIII, RUV4 implementation strategies for pseudobulked scRNA-seq and introduces RUVIII PBPS to control FDR without technical replicates or…
A systematic evaluation of highly variable gene selection methods for single-cell RNA-sequencing.
Benchmarks 47 highly variable gene selection methods across 19 scRNA-seq datasets and 18 criteria; proposes hybrid mixHVG method with an open-source R package.
Preprints
Comparative Analysis of Feature Selection Methods for Single-Cell RNA Sequencing Data
Benchmarks 16 feature selection methods on PBMC3K and Visium scRNA-seq; mean overlap 23.7%, 1,150 genes consistently selected.
COBRA: Cell-type-specific Orthogonal Batch effect Removal Algorithm in single cell RNA-sequencing data
COBRA is a linear-model batch correction method for scRNA-seq that adjusts cell-type-specific batch effects in the full gene space, with iterative…
Topic history
| Date | Rank |
|---|---|
| #15 | |
| #12 | |
| #39 | |
| #2 | |
| #5 | |
| #22 | |
| #38 | |
| #16 | |
| #25 | |
| #31 | |
| #10 | |
| #22 | |
| #16 | |
| #25 | |
| #31 | |
| #39 | |
| #2 | |
| #21 | |
| #8 | |
| #2 | |
| #13 | |
| #18 | |
| #9 | |
| #40 | |
| #23 | |
| #24 | |
| #5 | |
| #5 | |
| #7 | |
| #6 | |
| #9 |