Challenge 3 scoring metrics and 8%

Submitted by DataMiner on Mon, 2011-07-25 12:10

Hi, I have two (maybe related) questions:

1) Could you please announce (and not change them just before/after the deadline :) the scoring metrics for challenge 3? 

2) Can you write a bit more on: "The mCherry measurements as well as replicates of the tested promoters indicated that we can distinguish between any two promoters whose activities differ by as little as ~8%". What does this exactly mean (8% of what)? Will this affect the scoring metrics?

Thank you!

Ewa

‹ challenge1: perturbation experiments Challenge 4: Compensation ›

Comments

Scoring and 8% Error

Submitted by Dream_manager on Tue, 2011-07-26 15:40.

Hi Ewa, 

 

 

1) We have NOW uploaded a document describing the scoring, you can access here the link or  in the Challenge 3 webpage.
 
2) When stating that a certain promoter activity is for example 1, and the activity of another promoter is 1.2, one needs to know if this 20% difference between them is within the experimental error or not (If the activity difference is larger than the experimental error, it can then be attributed to their sequence differences).
 
We estimated the experimental error by having replicates for each promoter (biological and technical), and also from the mCherry measurements. Since all strains have the same promoter driving the mCherry, you expect the mCherry variance to represent the experimental error.
The typical standard error we had was 2% of the measured activity value.  Taking a confidence interval of +-2 standard errors, one can say that the promoter with activity of 1+-0.04 is significantly different from the promoter with the activity of 1.2+-0.048 and therefore in this case the differences in activities can be attributed to the differences between their sequences and not simply to the experimental error.
 
Thanks for your interest in this challenge
 
Pablo

Thank you so much for the

Submitted by DataMiner on Tue, 2011-07-26 22:02.

Thank you so much for the explanation. I am impressed with the scoring metrics, now I get why it took you some time to prepare it :)

all the best

Ewa

Hi again,   I would like to

Submitted by DataMiner on Wed, 2011-07-27 10:22.

Hi again,

 

I would like to ask about calculation of p_4. What exactly is "rank of proximity" gamma_ip ? I imagine this would be the rank of p in the order as measured by, for example root mean squared error. Am I correct, or is it measured in a different way?

 

Thank you dear dream-team!

Ewa

Hi Ewa  gamma_ip, the rank of

Submitted by Dream_manager on Wed, 2011-07-27 14:43.

Hi Ewa

 gamma_ip, the rank of proximity is just the distance so the absolute value of the rank difference which indeed is the RMS.

thanks again for your comment and we are delighted you appreciate our hard work

The Dream-team