roc
Calculates Receiver operating characteristic (ROC) curves for a given
set of networks with respect to a reference complex set.
Call the tool without any command line arguments to get detailed information about the parameters for this program.
See also:
Example calls
- calculate ROC curve based on MIPS reference set (
-ref
and -negref
)
- use localization data by Huh et al., 2003
- compare three different score networks (the last three parameters)
- plot up to a false positive rate of 0.01 (
-fpmax
)
- draw plot and save as png file (
-out
and -png
)
- we use a different set to determine true positives than for the false positive generation (see also: ROC curves)
- the MIPS dataset contains two large complexes (>50 proteins) which induce a lot of true positive edges (n*(n-1)/2) for each complex
- therefore there is a complex set where these two complexes are removed (
mips_s50.txt
) which is used for true positive determination
- for the negative set we use the full complex set (
mips_complexes.txt
)
./cmdline.sh roc -ref data/complexes/mips_s50.txt -fpmax 0.01 -loc
data/localization/huh_loc_070804.txt -negref
data/complexes/mips_complexes.txt -out roc.png -png
data/scores/bootstrap_combined.txt.gz data/scores/hart_scores.txt.gz
data/scores/pe_combined.txt.gz
- same procedure as above but this time we save the curve data into files having the prefix "rocdata_" (no
-png
switch)
cmdline.sh roc -ref data/complexes/mips_s50.txt
-fpmax 0.01
-loc
data/localization/huh_loc_070804.txt -negref
data/complexes/mips_complexes.txt -out rocdata_
data/scores/bootstrap_combined.txt.gz data/scores/hart_scores.txt.gz
data/scores/pe_combined.txt.gz