Why Use TAMALPAIS? And Other Methods Notes
Mark Bieda (revised 08/23/07)
The TAMALPAIS server is located here
(1) Why use TAMALPAIS?
The short answer: TAMALPAIS works!
The predictions from TAMALPAIS have been repeatedly verified in our lab and others.
In particular, the high stringency level (L1) gives ~95-100% verification rates routinely.
For a TABLE of actual results, see below:
This table is from Bieda et al, 2006 Genome Research 16:595 Supplementary Information and the full supplementary information is found here.
Supplementary Table 1: PCR Confirmations of Peak Predictions
"confirmation"
rate
total # incorrect % incorrect
Predicted Sites
Positives
L1 29 0 0.0% 100.0%
first appears at L2 6 0 0.0% 100.0%
first appears at L3 3 1 33.3% 66.7%
first appears at L4 5 4 80.0% 20.0%
Negatives 10 1 10.0% 90.0%
Individual Array
Positives
L1 82 4 4.9% 95.1%
first appears at L2 13 0 0.0% 100.0%
first appears at L3 7 1 14.3% 85.7%
first appears at L4 11 5 45.5% 54.5%
Negatives 36 8 22.2% 77.8%
Notes on the above table:
'Predicted Sites' refers to sites found in at least 2 of 3 replicates; these are the 'S50G2' sites from the browser.
Because we have good results with the L2 and L3 levels, we routinely combine the results from these levels;
these are supplied from the TAMALPAIS server as the L2L3combo files.
Note that every L1 peak is in the L2 set, but under some conditions, some L2 peaks are not in the L3 set.
(2) How were the gene localizations and gene identifications found?
Custom software was developed by Mark Bieda.
We downloaded files of KnownGenes for each assembly from the UCSC genome browser.
For hg16, you can reference Bieda et al, 2006 Genome Research 16:595
For hg17, the files were downloaded on April 14, 2006.
For hg18, we mapped the hg17 data to hg18 using the program liftOver (available at UCSC Genome Browser).
For mm5 and mm8, we downloaded files on August 12, 2007.
Files are available upon request.