Fork me on GitHub
Math for the people, by the people.

User login

fold change question

Primary tabs

fold change question

Hi,
I am trying to develop an estimate for the fold change in the expression level of a gene, under 2 conditions A and B. I take three signal intensity measurements in each condition, say, IntA(1), IntA(2), IntA(2) etc. The signal intensity provides a direct measure of expression level.
What is the best way to develop an estimate for the fold-change between A and B?
Two possible ways:
1)
IntA(mean)= Mean of the 3 intensity measurements under condition A.
IntB(mean)= Mean of the 3 intensity measurements under condition B.
Mean fold change = IntA(mean)/ IntB(mean)

2)
I calculate all 9 possible foldchanges: IntA(1)/IntB(2); IntA(1)/IntB(3); IntA(2)/IntB(1); IntA(2)/IntB(2) etc..
Then Mean fold change = Mean of the 9 fold changes.

Any feedback on which method is more suitable would be very much appreciated.
Thanks in advance!


thank you both for your replies.

OK, dh2718. Let Fa and Fb be the numerator and the denominator of the estimated fold change, F
I tried to minimize the following function:
D= [(IntA(1)-Fa)^2+(IntB(1)-Fa)^2+(IntA(1)-Fa)^2+(IntB(2)-Fa)^2+...]
Result: Fa= [IntA(1)+IntA(2)+IntA(3)]/3
Fb= [IntB(1)+IntB(2)+IntB(3)]/3

However, if I try to minimize:
D= [(IntA(1)-F*IntB(1))^2+(IntA(1)-F*IntB(2))^2+(IntA(1)-F*IntB(2))^2+...]

F= [IntA(1)/IntB(1)+IntA(1)/IntB(2)+........]/9

I will stick to the first minimization to estimate Fa and Fb.
Thanks for your input

Off hand I would recommend method 1.

Method 3 - The least mean square estimate. F being the meand fold change, minimize the F-function:

[IntA(1) - F*IntB(1)]^2 + [IntA(2) - F*IntB(2)]^2 + ...

Subscribe to Comments for "fold change question"