Graphics Reference
In-Depth Information
where T
is the total number of time points and RT
(
t
)
(
A, R, T
)
denotes the lack-of-
fit score of
with the tth sample deleted. From (
.
), the diagnostic function
can apparently screen out any gene that has one or more time points that deviate
greatlyfromthemodelin(
.
).Alarge Diag valuesuggeststhateithergeneexpression
levelsofthetripletatoneormoretimepointsmaybeproblematicorthatthetriplet
does not fit the response surface well. It is clear that the criteria that should be used
to check whether a triplet fits the response surface well are RT
(
A, R, T
)
(
A, R, T
)
C
and
Diag
C
,wheretheC
i
valuesareconstants.hefollowingscorefunction
was defined to measure the overall fitting:
(
A, R, T
)
Score
(
A, R, T
)=
RT
(
A, R, T
)[
+
Diag
(
A, R, T
)]
(
.
)
he SRS algorithm was applied to two sets of microarray data: GeneChip data on the
yeast cell cycle (Cho et al.,
) and cDNA microarray data on the yeast cell cycle
(Eisen et al.,
). here were
genes with
time points from the affymetrix
Ye
chip inthe formerset. Aterfiltering,
genes wereretained andprocessed
by the SRS algorithm. Among those triplets formed,
had RT scores of
<
.
and Diag
.
. hose best-scoring triplets that have biological meaning are plotted
in Fig.
.
(Fig.
in Xu et al.,
). A few targets in the network of Fig.
.
asso-
ciated with regulators either carry out similar cellular functions or are involved in
the same cellular process. For instance, CDC
is an essential gene for cell division
and DNA recombination; four regulators of CDC
(SMC
, NIP
, BTT
, NUM
)
are functionally related. HAP
is a transcription factor; the predicted repression of
HAP
on CYC
agrees with the literature that HAP
represses CYC
under anaer-
obic growth and activates CYC
under aerobic growth. FAA
and HES
, which are
relatedtocellular lipidmetabolism andergosterol biosynthesis, havebeenimplicated
in HAP
regulation in the literature.
Besides those described in Xu et al. (
) and Shieh et al. (
), there is a wide
range of sotware available for the visualization of gene networks; for instance, for vi-
sualizingbiomedicalnetworks(BioMiner,Graphviz,Ospray,amongothers),forinte-
grating genomics and proteomics data fromexternal databases tovisualize pathways
or genetic networks (Dynamic Signal Maps, KnowledgeEditor, PathFinder, Pathway
Assist, PubGene, Vector, PathBlazer and others), and for modeling and simulating
gene regulatory networks (Genetic Network Analyzer, GenePath, among others).
<
A Regression Approach
1.3.4
Recently, there have been a few studies on transcriptional compensation (TC) inter-
actions (Lesage et al.,
; Kafri et al.,
; Wong and Roth,
). Following the
loss of a gene, the expression of its compensatory gene increases; this phenomenon is
known as TC. Reverse-transcription (RT)-PCR experiments have shown that, aside
from TC, compensatory gene expression decreases in some cases following the ab-
sence ofa gene;we call this phenomenon transcriptional diminishment (TD).heTC
interactions among a group of
yeast genes, which are synthetic sick or lethal to
SGS
orRAD
(Tong et al.,
), are of interest (Shieh et al.,
). he SRS algo-
