Figure 3. Training a gapped kmer support vector machine (gkmSVM) classifier trained on zGPAEs.
(A) Pipeline for training and cross-validation of gkmSVM classifier on zebrafish periderm enhancer candidates. (B) Receiver Operating Characteristic (ROC) and Precision-Recall (PR) curves using the gkmSVM trained on zGPAEs. au, area under. Color of curves corresponds to SVM scores. (C) Violin plots showing SVM scores of zebrafish genome tiles with 0% or at least 90% overlapped with the training set (GPAEs). (D) Average H3K27Ac ChIP-seq reads at the 30,000 elements with the highest or lowest scores from the gkmSVM trained on zGPAEs. (E) GO enrichment assay for genes associated with the top-scoring tiles 10,000 tilesincluding those that overlap the training set.
Figure 3—source data 1. Density plot for H3K27Ac ChIP-seq reads, as plotted in Figure 3D.
elife-51325-fig3-data1.csv (2.2KB, csv)
Figure 3—source data 2. Barchart for GO enrichment assay, as plotted in Figure 3E.
elife-51325-fig3-data2.xlsx (34.3KB, xlsx)
Figure 3—figure supplement 1. GO enrichment assay of gene expression for the top-scoring 10 K tiles that do not overlap zGPAEs.
