Application of machine learning methods to histone methylation ChIP-Seq data reveals H4R3me2 globally represses gene expression

Table 7 MARS model knockout analysis

MARS model knockouts	log2 fold change (predicted WT/predicted KO)
H3K27me2	-0.742
H4R3me2	-0.506
H3K27me3	-0.244
H2BK5me1	-0.158
H3K79me2	-0.046
H3K4me3	0.054
H3K79me3	0.421
H4K20me1	0.715
H3K36me3	0.941
H3K27me2-H3K27me3	-2.333
H2BK5me1-H3K27me2	-1.329
H3K27me2-H4R3me2	-1.248
H3K27me2-H4K20me3	-0.973
H3K27me2-H3K79me2	-0.789
H3K36me3-H3K4me3	0.996
H3K36me3-H4R3me2	1.011
H3K79me3-H4K20me1	1.136
H3K36me3-H4K20me1	1.327
H3K36me3-H3K79me3	1.362
H3K79me1-H3K79me3	1.553

The log₂ fold changes (predicted WT/predicted KO) in average gene expression for single and double knockouts in the MARS model. In silico knockouts were performed by setting mark amplitudes to zero while fixing all other marks at their experimental values and making model predictions for each gene. The top 5 most repressive and 4 activating fold changes for single as well as the top 5 most repressive and activating double knockouts are shown. Rows are sorted according to log₂ fold change for single and double knockouts separately.

ISSN: 1471-2105