Skip to main content
Fig. 3 | Genome Biology

Fig. 3

From: DISC: a highly scalable and accurate inference of gene expression and structure for single-cell transcriptomes using semi-supervised deep learning

Fig. 3

Evaluation of imputation performance by FISH. a and b were based on the MELANOMA dataset. a Scatter plots of Gini coefficients of 19 overlapped genes between FISH and RAW/the imputation, RMSE scores are shown. Two genes with different dropout levels are also shown, WNT5A and SOX10, with dropout levels of 99.85 and 85.18%, respectively. b Scatter plot of WNT5A and SOX10 expression levels between FISH and the imputation/RAW. FF scores were calculated across 13,564 cells in FISH and 8498 cells in the other plots. c RMSE of Gini coefficients of the RAW/the imputation against FISH in MELANOMA (19 genes) and SSCORTEX (33 genes). d FF scores of gene-gene distributions of the RAW/the imputation against FISH in MELANOMA (81 gene pairs) and SSCORTEX (528 gene pairs). e CMD of gene co-expression of the RAW/the imputation against FISH in MELANOMA (81 gene pairs) and SSCORTEX (528 gene pairs)

Back to article page

Genome Biology

ISSN: 1474-760X

Contact us