Cyber-T: Statistical Analysis for Control vs. Experimental Data
Example Control+Experimental Data Set

Input File
File to upload ( Format expected )
Fields delimited by

Delete lines with NULL Labels

Problems with uploading ?

Comments
Please enter any text that you would like to have as a header for the analysis output

Please check if you want the standard header from Cyber-T in your files

Data Input

 Note: Please note that for the following fields, column numbering starts at 0. Row numbering starts at 1.
Separate the numbers by spaces, not commas or tabs.

Row number where data starts (if second row is where data starts, type 2, since row numbering starts at 1 - if left blank, row 1 is where data starts)
Label columns (names of genes or other identifying labels, e.g. 0)
Control data replicate columns (e.g. 1 2 3 5)
Experimental data replicate columns (e.g. 4 6 8 7)

Minimum non-zero replicates required (e.g. 3). If left blank, the number of values of experimental/ratio data will be used. What does this mean ?

Posterior Probability of Differential Expression (PPDE) Analysis

   Click this button if you want to perform a Posterior Probability of Differential Expression test. What is a PPDE test?

Choose on which column you want to compute the PPDE
Number of ß-fit Iterations (time to compute exponentially increases with beta value)

Low Value Handling
Values less than will be set to 0 and ignored in calculations. (Including negative values will cause the application to die. If you need to correct for negatives, use the option below)
Offset the values by the lowest value in the dataset, effectively right-shifting the entire dataset. Useful if you want to include negative values. This value will be noted in the output.

Parameters for the Bayesian Standard Deviation Estimation

What do these mean?

Choose a sliding window size for approximating the variance of your values. How many total samples around the point of interest will give you a satisfactory estimate of the local variance?
Enter a confidence value here that applies to the Baysian variance estimate that you set immediately above. A decent default would be '10' or about 3 times the number of replicates per treatment (if # replicates is small). If left blank, the whole Bayesian estimate analysis will be skipped.

This analysis may take SEVERAL MINUTES to run, depending on the server speed, load, number of genes analyzed, and replicates.
  Cyber-T Main   Help   Contact Us