Mechanism-based Clustering of Genome-wide RNA Levels: Roles of Transcription and Transcript-Degradation Rates - Clustering Challenges in Biological Network - page 241

tions of genes over 8 possible mechanisms (excluding Mechanism 9) of control-
ling mRNA level in cells(defined in the previous section) was performed using the
binomial test to make inferences about a proportion for response rate based on a
series of independent observations. A hypothesis test was performed to compare
the average trajectories (also called sequences, time series, or profiles) of two
groups of genes coding for glycolytic and respiratory mRNA molecules. Each
group generates two average profiles, labeled TL and TR (See Fig. 12.3a). Thus,
the null hypothesis states that “there is no difference between the two profiles as-
sociated with glycolysis and respiration”, and the alternative hypothesis states that
“There exists a significant difference between the two profiles”. In order to test
these hypotheses, a multivariate approach for repeated measures analysis was used
to calculate p -values. The p -value for the two trajectories shown in Fig. 12.3a is
less than 0.0001, indicating that there are significant differences between the av-
erage glycolytic and respiratory mRNA level trajectories. The p -value for the two
curves shown in Fig. 12.3b is 0.2292, indicating that there is no significant dif-
ference between the average trajectories of the glycolytic and respiratory mRNA
synthesis rates.

12.3.1. Calibration of TL Data

To convert the TL data expressed in arbitrary unit to the corresponding values
in absolute unit (mRNA molecules per cell), we utilized the reference mRNA
abundance data complied by Beyer et al. [5] based on 36 datasets reported in the
literature. Out of about 5,700 mRNA abundance data that these authors collected,
we selected a total of 59 glycolytic and respiratory mRNA abundance values and
plotted them against the corresponding TL data, leading to a linear regression line
satisfying the relation, n =0 . 1114 TL 0 − 7 . 220,where n is the number of mRNA

molecules per cell, and TL 0 is the mRNA level measured in arbitrary unit at
t =0. The correlation coefficient of the straight line was 0.94245. This equation
wasusedtoconvertallthe TL data (measured at five different time stamps) in
arbitrary unit into the corresponding values in absolute unit.
Examples of the time courses of the average changes in TL , TR ,andthe
transcript-degradation/transcription ( D/T ) ratios of glycolytic and respiratory
genes are shown in Fig. 12.3 and 12.4. The glycolytic genes considered here
include PDC2, PFK1, PFK2, ADH5, PDC6, LAT1, PDA1, ADH2, GLK1, PGM1,
ADH3, TPI1, GCR1 ,and PDB1 and the respiratory genes include QCR6, COX13,
COX9, NDI1, CYT1, COX8, COX12, SDH4, COX5A, COX4, CYC1, QCR2, QCR8 ,
and COR1 .