Computing Fold Change Using Z scores
1
0
Entering edit mode
6.7 years ago

Hi,

The post tag might be inappropriate as this post mainly deals with Protein Structure Data, however the concept is similar to analyzing Differentially Expressed Genes.

I have two N dimensional row vectors of raw scores - S1 and S2, where the scores correspond to B factors of residues in the same protein. The vectors have same number of elements, say n1. As the scores in the vectors are computed using different methods, I converted the raw scores into standardized scores (Z scores) yielding vectors Z1 and Z2. The intention is to compare the two vectors of B factors and check for increase/decrease in fluctuation at each position.

I used the concept of Fold Change to compute the increase/decrease in fluctuation at each residue position. However, I got stuck in situations where one Z score is +ve and the other is -ve; e.g. when Z1(r=5) > 0 and Z2(r=5)<0. One possible way I thought of doing this was converting the Z score into positive scale (by adding the absolute minmum of the two vectors to the Z score vector). However, the issue then would be that the original ratio will not be retained as mentioned below.

[Z1(r=5)] + |m| / [Z2(r=5)] + |m| is not always equal to Z1/Z2. 'm' is the minimum value of the 2 vectors.

Any suggestions?

Thanks,
Sambit.

RNA-Seq • 3.1k views
0
Entering edit mode
why don't you compute the fold change first and then standardize that quantity?
0
Entering edit mode
6.7 years ago
vassialk ▴ 200

Use Expander software, everything for M arrays is implemented there.