Biology Reference
In-Depth Information
Table 4.2 Counts of the various combinations of zero and nonzero
measurements for those signatures that exhibit at least
one nonzero measurement in the four basal MPSS runs
(B1:1, B1:2, B2:1, and B2:2)
Basal 1 (B1) and
Nonzeros in B1
One zero in B1
Two zeros in B1
Basal 2 (B2)
(6875)
(4487)
(8209)
Nonzeros in B2 (10,090)
5239
1558
3293
One zero in B2 (6839)
817
1106
4916
Two zeros in B2 (2642)
819
1823
0
probabilities of the zero count in B2 given the zero count in B1 in table 4.3.
The table for the conditional probability of the zero count in
B1 given the zero count in B2 has roughly the same characteristics as
table 4.3. It is clear that having nonzeros in both MPSS runs in B1 is
predictive of having nonzeros in B2. However, having one zero or
two zeros in B1 does not clearly determine what to expect in B2.
The data leads to the conjecture (which can be refined by address-
ing sequence-specific characteristics) that signatures with count
numbers below some threshold have a higher probability of not being
measured by the two MPSS runs (B1:1 and B1:2 in the B1 case) and,
thus, their count in one of the MPSS runs is zero. In these instances,
the count of zero does not have to be considered as a real count (i.e., as
a possible measurement value given the actual count), but rather as
an artifact to be interpreted as a failure in measuring a signature whose
value is not necessarily small.
It is straightforward to test a necessary condition that follows
from the above conjecture. Consider the signatures whose counts in
B1 had exactly one zero in one of the two MPSS runs, but whose count
in B2 had two nonzeros (such as, for example, signature 2 in table 4.1).
Measure the distribution of counts for those signatures in B2 on the
one hand, and in B1 (only the nonzero measurement) on the other. If
the measurement of zero in one of the two MPSS runs in B1 were due
to the fact that these signatures just did not go through the whole
process, then one should expect that the distribution of the counts of
Table 4.3 Data from table 4.2 posed in terms of conditional probabilities of
the zero count in B2 given the zero count in B1
Prob(B2 | B1)
Nonzeros in B1
One zero in B1
Two zeros in B1
Nonzeros in B2
76.2%
34.7%
40.1%
One zero in B2
11.9%
24.6%
59.9%
Two zeros in B2
11.9%
40.6%
 
Search WWH ::




Custom Search