Biology Reference
In-Depth Information
hybridization noise dq
hyb
.
For replicate pairs in group
G
1
and
G
2
, the noise
can be expressed respectively as: dq
1
dq
hyb
.
Assuming the two sources of noise are independent of each other, their
variances can be obtained by s
hyb
=
=
dq
hyb
, dq
2
=
dq
prep
+
dq
hyb
=
s
1
, s
prep
=
dq
prep
=
s
2
−
s
1
,
where s
1,2
can be computed from eq. (1).
In figure 4.4 is shown s
1
(q
0
) (dotted line) and s
2
(q
0
) (solid line)
versus the expected value of the expression level q
0
. Although the dif-
ference between s
2
and s
1
is small in comparison with s
2
, s
1
(q
0
) is
consistently smaller than s
2
(q
0
) for all the values of q
0
≥
1. This should
be so because the difference between s
2
and s
1
accounts for the sample
preparation noise: this difference, albeit small, is real.
The dependence of s
prep
versus q
0
is plotted in the inset of figure 4.4.
One finds that the dependence of s
prep
on the expression level q
0
can be
well approximated by:
−
q
2
−
3
σ
prep
≈× +
7 2
.
10
0 045
.
×
10
(3)
0
The constant first term dominates the sample preparation noise for
expression values q
0
≥ 2.
Figure 4.4
The dependence of the noise strength s
on the expected values of
the gene expression for replicates in groups
G
1
and
G
2
. (inset) The variance of
the sample preparation noise s
prep
=
1,2
s
2
)
1/ 2
is shown; s
prep
has very weak
dependence on the expression value for the large expression levels q
(s
2
−
0
≥ 2
, and
−
θ
(.
7 2
×
10
3
+
0 045
.
×
10
)
can be fitted by
−
0.5
for q
0
≥ 1.
0
Search WWH ::
Custom Search