MICROARRAY DATA ANALYSIS IN AFFYMETRIX GENE CHIP - Quantitative Medical Data Analysis Using Mathematical Tools and Statistical Techniques - page 144

P (i)

gene;E represents a gene level threshold, and I is an indicator function.
The two thresholds, P (i)

probe;E and P (i)

, are pre-specied constants. The

gene;E

probe level threshold P (i)

probe;E provides a cuto for probe discrepancy be-
tween two groups, and the gene level threshold P (i)
gene;E

provides a cuto

for determining dierential expressions.

2.2.4. Probe Weighted Rank Approach

The probe rank has limitations for detecting expression dierences for genes
with extremely high intensity (e.g. in the 98th100th percentile). For genes
within this range, their ranks tend to be similar. It becomes dicult to iden-
tify altered genes in this range because of a very small rank dierence. In
practice, this situation is rare and likely occurs in a highly abundant gene.
The gene intensity is likely in the range of high percentile at each experimen-
tal group (i.e., treatment and control). Though the dierence of intensity
between the two experimental groups may be substantial (e.g., > 2 fold for
treatment versus control), the rank dierence remains relatively small. To
overcome the problem, we introduce a weighting factor to the rank. By giv-
ing the weighting factor, a gene with high intensity is likely to have its rank
score amplied substantially because the weighting factor is proportional
to the probe intensity. As a result, dierence of the weighted rank scores
becomes large between treatment and control groups. Mathematically, the
weighted rank approach is the same as the rank approach except Y i;j;k
multiplied by a weight w i;j;k . The weight w i;j;k for the jth probe in the

ith gene on array k, is dened as log 2 (P M i;j;k )= P i=1

P J i
j=1 log 2 (P M i;j;k )
where P M i;j;k is PM intensity for the jth probe in the ith gene of array
k, m is the total number of genes in an array, and J i represents the total
number of probes for gene i. The PM weighted rank Y (weighted)
i;j;k

becomes

Y i;j;k w i;j;k . The probe weighted rank approach uses Y (weighted)

to com-

i;j;k

pute percentile dierence (i.e., D (i;j)
A;B

and D (i;j)

B;A ) and select dierentially
expressed genes. By giving a larger weight on high intensity probes, the
probe weighted rank approach can increase the power of detecting expres-
sion dierences for highly abundant genes better than the rank approach
does.

2.3. An Integrated Bioinformatics Tool

Analysis of gene selection often yields a long list of genes with detailed
information, such as gene expression fold change, p value, and numerous