Since there are many genes that only have low-PCC coexpressed gene pairs, but they are functionally relevant. To keep the weak but significant gene coexpression from being discarded, we introduced a new measure of gene coexpression, Mutual Rank (MR). In our assessment system, MR shows higher performance to predict gene function than PCC. (see
coexpression versions)
Here,
Rab indicates correlation rank from gene A to gene B.
Example: Top 5 coexpressed gene list for gene A
| Rank |
PCC |
Locus |
| 0 |
1.00 |
gene A |
| 1 |
0.93 |
gene E |
| 2 |
0.93 |
gene D |
| 3 |
0.93 |
gene B |
| 4 |
0.93 |
gene C |
| 5 |
0.92 |
gene F |
In this case,
Rab = 3.
We do not introduce any cut-off threshold. Namely even if the correlation value between two loci is insignificantly low, the ranks are numbered from the strongest pair to the weakest pair.
Correlation rank is asymmetric, namely the rank of gene B from gene A is not the same as the rank of gene A from gene B.
And thus, those two ranks are geometrically averaged, which we call Mutual Rank (MR).
|
MR(AB) = √ (Rank(A→B) x Rank(B→A))
|
