Skip to main content
Fig. 3 | Genome Biology

Fig. 3

From: Integration of single-cell multi-omics data by regression analysis on unpaired observations

Fig. 3

UnpairReg gene expression prediction is consistent with the paired data. A Cell-level PCC between the predicted and multiome gene expression. The x-axis represents the cell-level PCC between UnapirReg predicted gene expression and multiome gene expression, while the y-axis gives the cell-level PCC between GAS and multiome data. Color in A to C indicates the different methods. Orange represents for UnpairReg, and green represents GAS. B Gene-level PCC between predicted and multiome gene expression. The x-axis represents the gene-level PCC between UnapirReg predicted gene expression and multiome gene expression, while the y-axis represents the gene level PCC of GAS and multiome data. C Gene expression of multiome and UnpairReg for one gene. R2 is the r-squared as a goodness-of-fit measure for the linear regression model. p-value is for the F-test of linear regression. The y-axis and x-axis give predicted gene expression and log10 (1 + count) in a cell. D Alignment error of predicted gene expression. For each cell, represented by the predicted gene expression vector, we compute its distance with all cells (observed gene expression). Alignment error for a cell is defined as the number of cells that have a closer distance than the true match (the same cell). E The performance metrics AUPR for UnpairReg cis-regulatory coefficients and other methods. The ground truth is the variant-gene links from eQTLGen. We divide RE-TG pairs into different groups based on the distance of RE and the TSS of TG. There are 8634, 5791, 7431, 21,282, and 33,205 RE-TG pairs in 0–3k, 3–10k, 10–20k, 20–50k, 50–100k, and 100–150k, respectively. PCC denotes the Pearson’s correlation coefficient of RE promoter. Distance denotes the decay function of the distance to the TSS; random denotes the uniform distribution. This figure corresponds to the PMBC data

Back to article page