Reconstruction Set Test (RESET): A computationally efficient method for single sample gene set testing based on randomized reduced rank reconstruction error

Abstract

We have developed a new, and analytically novel, single sample gene set testing method called Reconstruction Set Test (RESET). RESET quantifies gene set importance based on the ability of set genes to reconstruct values for all measured genes. RESET is realized using a computationally efficient randomized reduced rank reconstruction algorithm (available via the RESET R package on CRAN) that can effectively detect patterns of differential abundance and differential correlation for self-contained and competitive scenarios. As demonstrated using real and simulated scRNA-seq data, RESET provides superior performance at a lower computational cost relative to other single sample approaches.

Fig 1. Workflow representation of the randomized version of RESET as defined by Algorithm 3.

Fig 2. Classification performance of RESET.det, RESET.ran, VAM, GSVA, ssGSEA, and PLAGE on scRNA-seq data simulated according to the block design for a single gene set as detailed in the Methods section in S1 Text.

Each panel illustrates the relationship between the area under the receiver operating characteristic curve (AUC) and one of the simulation parameters. The vertical dotted lines mark the default parameter value used in the other panels. Error bars represent the standard error of the mean.

Fig 3. Classification performance of RESET.det, RESET.ran, VAM, GSVA, ssGSEA, and PLAGE on scRNA-seq data simulated according to the complex design for a single gene set as detailed in the Methods section of S1 Text.

Each panel illustrates the relationship between the area under the receiver operating characteristic curve (AUC) and one of the simulation parameters. The vertical dotted lines mark the default parameter value used in the other panels. Error bars represent the standard error of the mean.

Fig 4. Classification performance of RESET.det, RESET.ran, VAM, GSVA, ssGSEA, and PLAGE on scRNA-seq data simulated according to the pure self-contained design for a single gene set as detailed in the Methods section in S1 Text.

Each panel illustrates the relationship between the area under the receiver operating characteristic curve (AUC) and one of the simulation parameters. The vertical dotted lines mark the default parameter value used in the other panels. Error bars represent the standard error of the mean.

Fig 5. Classification performance of RESET.det, RESET.ran, RESET.pervar, VAM, GSVA, ssGSEA, and PLAGE on scRNA-seq data simulated according to the pure competitive design for disjoint and equally sized genes as detailed in the Methods section in S1 Text.

Note that RESET.det and RESET.pervar have identical performance in this case since the gene sets all have the same size. Each panel illustrates the relationship between the area under the receiver operating characteristic curve (AUC) and one of the simulation parameters. The vertical dotted lines mark the default parameter value used in the other panels. Error bars represent the standard error of the mean.

Fig 6. Overall classification performance of RESET.det, RESET.ran and RESET.pervar on scRNA-seq data simulated according to the block design with disjoint and equal sized gene sets as detailed in the Methods section in S1 Text.

Each panel illustrates the relationship between the area under the receiver operating characteristic curve (AUC) and one of the simulation parameters. The vertical dotted lines mark the default parameter value used in the other panels. Error bars represent the standard error of the mean.

Fig 7. Visualization of cell type BioCarta pathway enrichment as computed using either VAM or RESET scores on the PBMC scRNA-seq data.