how define a value for features with different frequency?

Question

[star]

I have a table like below 1), that contains different histone marks (with different numbers for each mark) from different studies. I like to have a table like 2), and in fact ingrate all similar marks together as one column, but I do not know which measurement is more accurate (mean/median/percentile or ...).

The value for each mark is based on the number of reads that extracted from BAM file and They were normalized based on RPKM method sample.

1)

      H3k4me1-study1   H3k4me1-study2  H3k27ac-study1  H3k27ac-study2   H3k27ac-study3
A           5                 7              10               4               25
B           12                6              8                6               18
C           9                 5              9                3               15

2)

      total-H3k4me1  total-H3k27ac
A     
B
C