Page 317 - Read Online
P. 317
Shen et al. Evaluation of microRNAs normalization approaches
[33]
[32]
Two statistical algorithms (geNorm and NormFinder ) together had the lowest systematic error (stability
were used to estimate the stabilities of the miRNAs value) of 0.133 by NormFinder [Supplementary Table 3],
profiles. The algorithm of geNorm is to calculate suggesting their good stabilities for normalization.
the average pairwise variation (V) for each miRNA If we included the means of global miRNAs in
with all others across the samples, and estimate the stability analyses, the M score and stability
a stability score (M) defined as the average V of a value for miR-30c/mean of miRNAs combination
miRNA with all others. The less stable miRNA with were, respectively 0.022 and 0.088, lower than the
the highest M is gradually removed until the 2 most combination of miR-30c with miR-30b or miR-126
stable normalizers are obtained. The algorithm of [Supplementary Table 3]. This suggests that the mean
NormFinder is to calculate the inter- and intra-group of global miRNAs may be a good normalizer due to
variances of the log-transformed miRNAs expression its high stability among samples. The stabilities of the
data, and integrate it into a stability value to represent manufacturer-recommended normalizer ncRNAs (U6
the systematic error of each miRNA. A lower value of snRNA, RNU44 and RNU48) ranked much lower in
systematic error indicates a more stable miRNA, and the 17th to 140th range out of a total 157 candidates,
the combination of the most stable miRNAs is selected indicating their poor stability in HCC tissue.
as the normalizer. Different types of normalizers (array
recommended ncRNAs, mean of global miRNAs and Aberrant miRNA panels identified by using
the most stable miRNAs combination) were separately varied normalizers
used to generate miRNA expression profiles from After excluding non-abundant miRNAs and those with
HCC tissue samples for future statistical data analysis. missing data in over 50% of samples, a total of 361
miRNAs were finally analyzed. Using 3 endogenous
Before performing any statistical analysis, the genome- controls (U6 snRNA, RNU44 and RNU48) as the
wide miRNA profiles were checked to ensure the normalizer, we found 46 miRNAs significantly
reliability and abundance of miRNAs. If the missing dysregulated (P < 0.001) in HCC tumor tissues with
data (Cq ≥ 40) for any miRNA exceeded 50% of at least 2-fold changes in expression [Supplementary
samples, this miRNA was excluded from further data Table 4, Figure 1A]. Most miRNAs (43) were significantly
analyses. Paired t-test was used to identify miRNAs down-regulated in HCC tumor tissue (from 2 to 10-fold),
that were significantly different by the univariate test and only 3 miRNAs were significantly up-regulated with
(P < 0.001) with at least a 2-fold expression change fold changes of 5 to 9.
between paired HCC tumor/non-tumor tissues or
HCC cases and matched controls. Volcano plots were Using the mean of all miRNAs as a normalizer, a total
generated to describe the distribution of significant of 26 miRNAs were significantly different between
miRNAs with over 2-fold changes. Hierarchical HCC tumor and non-tumor tissues with over 2-fold
clustering and heat maps were produced with average changes [Table 1, Figure 1B]. The aberrant expression
linkage and Pearson correlations to examine the pattern was quite different from that identified by using
classification of samples based on significant miRNAs. 3 endogenous controls. More miRNAs (17) were
All statistical analyses were performed using BRB- upregulated 2- to 16-fold compared with 9 significantly
ArrayTools (version 4.4) developed by Dr. Richard downregulated miRNAs (2-5 fold) in HCC tumor
Simon and the BRB-ArrayTools Development Team tissues. The expression levels of endogenous controls
(http://linus.nci.nih.gov/BRB-ArrayTools.html) [34] and (U6 snRNA and RNU44) were increased while RNU48
Statistical Analysis System 9.0 (SAS Institute). The was reduced in tumor tissue when using the mean
panels of significant miRNAs identified by different of all miRNAs as a normalizer, but no significant
normalizers were compared using a web-based difference was obtained (data not shown). The fold-
InteractiVenn tool (http://www.interactivenn.net/) [35] to changes between tumor and non-tumor tissues were
determine the consistent and discrepant miRNAs. varied from -1.28 to 4.09 times.
RESULTS Supplementary Table 5 and Figure 1C display a panel
of 17 miRNAs aberrantly expressed in HCC tumor
The stabilities of global miRNAs/ncRNAs tissue using 2 stable miRNAs (miR-30c and miR-
Using geNorm and NormFinder tools, we separately 30b) identified by the geNorm tool as the normalizer.
examined the stabilities of global miRNA profiles in Six were significantly upregulated 2- to 15-fold;
HCC tissues. Among 157 miRNAs/ncRNAs tested 11 were downregulation 3- to 6-fold in HCC tumor
in HCC tissues, the combination of miR-30c/miR- tissue. Similarly, using 2 stable miRNAs (miR-30c
30b had the smallest M score of 0.024 by geNorm and miR-126) identified by the NormFinder tool as
[Supplementary Figure 1], and miR-30c/miR-126 the normalizer, we found 20 significantly deregulated
308 Hepatoma Research ¦ Volume 2 ¦ November 18, 2016