Skip to main content

Table 2 MARS model terms

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

Coefficient

Hinge function

Genes with non-Zero value

5.531222

1

17635

-10.31971

h(H3K27me2-0.0611382)

6892

1.325129

h(H3K79me3-0.0948497)

7218

-10.70662

h(0.0948497-H3K79me3)

10417

-58.18392

h(0.0611382-H3K27me2)

10743

-3.112329

h(0.559645-H4K20me1)

13551

-3.327909

h(0.545052-H3K36me3)

16872

15.2676

h(0.125391-H4R3me2)

17316

118.4095

h(H3K27me2-0.0611382)*h(H3K9me1-0.674218)

67

39.42698

h(H2BK5me1-1.49934)*h(0.0611382-H3K27me2)

106

-1.089521

h(0.545052-H3K36me3)*h(H4K20me1-0.673666)

3136

0.7436266

h(1.24429-H3K79me2)*h(H4K20me1-0.559645)

4055

47.71129

h(H3K79me1-0.055087)*h(0.0948497-H3K79me3)

4268

-258.7383

h(0.055087-H3K79me1)*h(0.0948497-H3K79me3)

6149

16.1397

h(H3K27me2-0.0611382)*h(0.674218-H3K9me1)

6825

1390.367

h(0.0611382-H3K27me2)*h(0.0913244-H3K27me3)

9906

3.235207

h(0.545052-H3K36me3)*h(0.673666-H4K20me1)

13736

-37.13981

h(0.438075-H3K36me3)*h(0.125391-H4R3me2)

15588

5.809395

h(1.49934-H2BK5me1)*h(0.0611382-H3K27me2)*h(H4K20me3-0.477337)

100

-100.9374

h(0.0611382-H3K27me2)*h(0.0913244-H3K27me3)*h(H4K20me1-3.93081)

203

-0.2791515

h(1.59266-H3K4me3)*h(1.24429-H3K79me2)*h(H4K20me1-0.559645)

2971

74.37772

h(0.545052-H3K36me3)*h(0.0625376-H3K79me1)*h(0.673666-H4K20me1)

6329

-237.028

h(0.0611382-H3K27me2)*h(0.0913244-H3K27me3)*h(3.93081-H4K20me1)

9703

90.39805

h(1.49934-H2BK5me1)*h(0.0611382-H3K27me2)*h(0.477337-H4K20me3)

10537

  1. Coefficients and hinge functions within a row are multiplied and added to the products of other rows to form the MARS model. The rightmost column indicates the number of genes for which the hinge function takes a non-zero value. Terms appear in the order in which they were built into the model by the greedy algorithm.