Skip to main content
Fig. 5 | Genome Biology

Fig. 5

From: Avocado: a multi-scale deep tensor factorization method learns a latent representation of the human epigenome

Fig. 5

The performance of each feature set when used for genomic prediction tasks. In each task, a supervised machine learning model is evaluated separately for each cell type using a 20-fold cross-validation strategy, with the mean average precision reported and standard error of the mean shown in the error bars. Each task considers only genomic loci in chromosomes 1 through 22. The tasks are predicting (a) expressed genes, (b) promoter-enhancer interactions, (c) replication timing, and (d) FIREs. In panel a, the coloring corresponds to the standard error with the mean average precision lying in the middle, whereas in the other panels the mean average precision is shown as the colored bar with standard error shown in black error bars. The statistical significances of differences observed in this figure are assessed in Additional fileĀ 2: Tables S2-5

Back to article page