October 25, 2005

README file for
"Combining Biological Networks to Predict Genetic Interactions"
Wong et al, PNAS 2004

The gene-pair data used to predict genetic interactions is contained in 27 files,
plus one file that lists the gene-pair characteristics we used.


A.  GENE-PAIR CHARACTERISTICS FILE: characteristics.txt
This file contains a heirarchical list of gene-pair characteristics.


File format:
<descripton of gene-pair characteristic> ; <identification number we assigned to characteristic>


For each root level characteristic, the name of file enclosed in parenthesis contains gene pairs that posses
that characteristic.
Abbreviated names for gene-pair characteristics are enclosed in curly braces.





B.  GENE-PAIR FILES: files ending in "_pairs" or "_prs"
These files contain gene pairs possessing the characteristic described by the respective file name.

File format:
<SGD indentification number of first gene>\t<SGD identification number of second gene>\t<identification number of gene-pair characteristic>


SGD = Saccharomyces Genome Database (http://www.yeastgenome.org/)



c.  NOTES
1. Treat "_pairs" files the same as "_prs" files.
2. In the characteristics.txt file, ignore lines beginning with "#".  
Those characteristics were ignored because they were 
	- redundant with another characteristic (usually the one listed immediately above), or 
	- too non-specific
3. A gene pair that possesses a characteristic also possess all corresponding parent characteristics.
4. When predicting SSL gene pairs, SSL-containg "2-hop" characteristics must be generated on the fly, according to gene pairs in the training set (i.e. A gene pair (or edge) that is not in the training set can not give rise to "2-hop" data used to train the model.).
5. Some SGD identification numbers have now been deleted or merged with other SGD identification numbers.  Pairs containing these retired SGD identification numbers should be removed when making future predictions.
6. Since these files were generated, the Saccharomyces Genome Database modified
its gene identification numbers.  Specifically, two zeros were inserted into
each identification number, such that "S0001234" became "S000001234".