How can we measure 2D Gel Data Quality?
Is it when the data is reproducible?
=
=
Reproducibility is generally measured by calculating the mean CV.
Low mean CV is often interpreted as reproducible data.
So, can mean CV be used as a metric for 2D gel data quality?
Consider:
Experiment
Measurements
Data
gel running
image analysis
data analysis
and
Consider:
CV
=
standard deviation
mean
but what if...
... we add a constant value to all our measurements?

... we don't do any background subtraction?

... we have amalgamation of spots?
CV
=
standard deviation
mean
the mean goes up
the CV goes down
We would get LOWER CVs, and interpret it as HIGHER reproducibility, although we have LOWER data quality!
After all, we don't want image analysis errors to mask the true variance in the experiment
low variance but incorrect
higher variance but correct
This highlights the need to assess the correctness of the measurements themselves!
How do we do that?
The good news is that in 2-DE you can LOOK at the measurements 
The bad news is that there are tens of thousands of measurements 
So what do you do when you have too many measurements to assess? 
You randomize and sample!
And we can write a software to do this for us
Then we define visual criteria for evaluating correctness
Spot detection evaluation criteria
Spot matching evaluation criteria
So by randomizing, sampling, and applying visual criteria for evaluation of correctness, we can generate an estimate of the overall correctness of the data. 
This is what this initiative aims to do!
A free tool for the scientific community!
...the measurements are wrong!?
What if...
...to help estimate the correctness of the measurements from 2D gel image analysis.
www.ludesi.com/free-tools
Gel IQ