Preservation of ranking order in the expression of human Housekeeping genes.

Abstract

Housekeeping (HK) genes fulfill the basic needs for a cell to survive and function properly. Their ubiquitous expression, originally thought to be constant, can vary from tissue to tissue, but this variation remains largely uncharacterized and it could not be explained by previously identified properties of HK genes such as short gene length and high GC content. By analyzing microarray expression data for human genes, we uncovered a previously unnoted characteristic of HK gene expression, namely that the ranking order of their expression levels tends to be preserved from one tissue to another. Further analysis by tensor product decomposition and pathway stratification identified three main factors of the observed ranking preservation, namely that, compared to those of non-HK (NHK) genes, the expression levels of HK genes show a greater degree of dispersion (less overlap), stableness (a smaller variation in expression between tissues), and correlation of expression. Our results shed light on regulatory mechanisms of HK gene expression that are probably different for different HK genes or pathways, but are consistent and coordinated in different tissues.

Full text

Page 1

Preservation of Ranking Order in the Expression of
Human Housekeeping Genes
Grace T. W. Shaw1,2, Edward S. C. Shih2,4,5, Chun-Houh Chen3, Ming-Jing Hwang1,2,4*
1Institute of Biomedical Informatics, National Yang-Ming University, Taipei, Taiwan, 2Institute of Biomedical Sciences, Academia Sinica, Taipei, Taiwan, 3Institute of
Statistical Science, Academia Sinica, Taipei, Taiwan, 4Chemical Biology and Molecular Biophysics Program, Taiwan International Graduate Program, Institute of Biological
Chemistry, Academia Sinica, Taipei, Taiwan, 5Institute of Bioinformatics and Structural Biology, National Tsing Hua University, Hsinchu, Taiwan
Abstract
Housekeeping (HK) genes fulfill the basic needs for a cell to survive and function properly. Their ubiquitous expression,
originally thought to be constant, can vary from tissue to tissue, but this variation remains largely uncharacterized and it
could not be explained by previously identified properties of HK genes such as short gene length and high GC content. By
analyzing microarray expression data for human genes, we uncovered a previously unnoted characteristic of HK gene
expression, namely that the ranking order of their expression levels tends to be preserved from one tissue to another.
Further analysis by tensor product decomposition and pathway stratification identified three main factors of the observed
ranking preservation, namely that, compared to those of non-HK (NHK) genes, the expression levels of HK genes show a
greater degree of dispersion (less overlap), stableness (a smaller variation in expression between tissues), and correlation of
expression. Our results shed light on regulatory mechanisms of HK gene expression that are probably different for different
HK genes or pathways, but are consistent and coordinated in different tissues.
Citation: Shaw GTW, Shih ESC, Chen C-H, Hwang M-J (2011) Preservation of Ranking Order in the Expression of Human Housekeeping Genes. PLoS ONE 6(12):
e29314. doi:10.1371/journal.pone.0029314
Editor: Tim Thomas, The Walter and Eliza Hall of Medical Research, Australia
Received August 18, 2011; Accepted November 24, 2011; Published December 22, 2011
Copyright: ? 2011 Shaw et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported in part by a grant from the National Science Council of Taiwan (NSC grant no. 97-2627-P001-004). No additional external
funding received for this study. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: mjhwang@ibms.sinica.edu.tw
Introduction
Housekeeping (HK) genes are defined as genes that are
permanently activated throughout the life cycle of the cell [1].
As they constitute the basic transcriptome for maintaining cellular
functions for cell survival, HK genes are also called maintenance
genes [2]. In general, genes that participate in essential cellular
processes can be considered to have HK functions. These include
genes involved in transcription [3], translation [4,5], energy
production and transmission [6,7], and maintaining cell organi-
zation, shape, and motility [8]. HK genes were initially discovered
in experiments involving RNA blot hybridization [9] and
immunological detection [4], when certain genes were found to
be expressed not only constitutively, but also at fairly constant
levels under all conditions tested [9]. On the basis of this stable and
ubiquitous expression, they have frequently been used as
endogenous references for various mRNA quantification experi-
ments [10–12]. However, studies have shown that the expression
of the HK genes actually fluctuates from tissue to tissue and often
from person to person [13,14]. Furthermore, in disease states, such
as liver and breast tumors, HK genes can exhibit very different
expression patterns from those observed in normal tissues [15].
While the assumption of constant expression may not be valid,
HK genes are useful references so long as their expression patterns
are characterized under the same conditions as those in which the
experiments are conducted [16,17]. For example, a stable
expression ratio of two HK genes [18,19] and stable mean value
of the expression of several HK genes [11,20,21] have been
proposed as internal controls in mRNA quantification experi-
ments. However, these propositions are not without flaws, because,
for example, the expression ratio of the RPL32 and GAPDH
transcripts, two commonly used internal controls in RNase
protection assays, is found to fluctuate in mitogen-stimulated
mononuclear cells [18]; likewise, a stable mean expression of
multiple HK genes in breast tumors is no guarantee that it will be
stable in other tissue types [21]. To control for these context-
dependent effects, in-advance characterization of HK genes is
required, but these characterizations are laborious and time-
consuming, so the possibility of finding other common properties
of HK genes is of significant interest.
In surveying several large-scale transcriptomics studies in the
literature, we noticed that HK genes seemed to follow a similar
ranking order in terms of their level of expression in different
tissues, even though the actual level might vary from one tissue to
another. For example, using data from a study report by Lisowski
et al. [22], which investigated stability of gene expression in cattle
tissues, we found that, although the expression of six common HK
genes, ACTB, GAPDH, HPRT1, SDHA, TBP, and YWHAZ,
differed in the kidney, liver, pituitary, and thyroid, the same
ranking order of level of expression was seen in all four tissues
(Figure S1). In the present study, by performing a statistical
analysis of microarray expression data for human genes, we
substantiated this observation and showed that an expert-curated
set of human HK genes indeed tended to exhibit a preserved
tissue-wide expression ranking. Furthermore, we identified the
main factors responsible for the preserved ranking and discussed
possible underlying mechanisms.
PLoS ONE | www.plosone.org1December 2011 | Volume 6 | Issue 12 | e29314

Page 2

Results
As detailed in the Methods, based on a manual curation for HK
genes [23] and an index for tissue specificity for TS genes [24], we
divided the human genes into the two sets of HK and non-HK
(NHK) genes, and from the NHK set we selected tissue-specific
(TS) genes and assigned the rest as middle-ranged (MR) genes. For
the Affymetrix’s GSE2361 dataset (Methods and Table S1) used to
illustrate the analysis below, this resulted in 388 HK genes and
12,687 NHK genes, and of the NHK genes 734 were TS genes
and 11,953 were MR genes. We then computed Kendall’s tau [25]
for each of the three (HK, MR and TS) gene sets to measure and
compare the extent to which the ranking order of gene expression
was preserved across tissues. Kendall’s tau (t) can be computed
either on pairs of tissues (? t ttt0) or on pairs of genes (? t tgg0); the two are
often used interchangeably below, since, in our case, they were
essentially identical (see Methods).
Preservation of expression ranking
Figure 1 shows that, as measured by Kendall’s tau, the
expression ranking order of HK genes in the 36 human tissues
of the GSE2361 dataset (Methods and Table S1) was more
concordant than those of the MR and TS genes. The ? t ttt0 for all
pairs of tissues sampled was 0.77 for HK genes, 0.59 for MR
genes, and 0.41 for TS genes (Figure 1; Table 1). The non-zero
Kendall’s tau for the MR genes and, to a lesser extent, for the TS
genes suggests some degree of preserved ranking in the expression
of the selected genes; in fact, even a randomly sampled set of genes
will exhibit a non-zero Kendall’s tau (Figure 2A). In general, there
was a significant correlation between the expression levels of the
same gene in any two tissues, because genes with high expression
levels, which are more likely to rank high than low, in one tissue
tend to have high expression levels in another tissue (e.g.
Figure 2B). Consequently, only when rankings were randomly
assigned were truly random (close to zero) Kendall’s taus produced
(Figure 2A). Compared to MR genes or a randomly selected group
of genes, the expression ranking of TS genes varied more between
tissues, i.e. producing a smaller ? t ttt0, owing to their expression in a
specific tissue and no, or little, expression in most other tissues
[24].
Similar results using the same analysis procedures were obtained
using two other Affymetrix and one Illumina datasets of human
gene expression (Table 1). Thus, in all the three Affymetrix
nucleotide microarray datasets and the one Illumina’s next-
generation RNA sequencing dataset analyzed (Table S1), it was
evident that the expression ranking of the selected HK genes
across tissues was much more preserved than would be expected if
the genes were picked randomly or if the genes were in the NHK
set (either the MR or TS set), although the mean value of their
Kendall’s taus can differ in different expression datasets (Table 1).
Contributions of the three factors of co-expression,
stableness, and dispersion
To investigate what produced the observed preservation of
expression ranking of the three gene sets, we considered three
factors, co-expression, stableness, and dispersion (Figure 3) and
carried out a tensor product analysis of a gene pair’s Kendall’s tau,
? t tgg0, as described in the Methods. Intuitively, when the expression
of two genes is highly correlated, the order of their expression
levels across tissues will be preserved (Figure 3A). Additionally,
when the expression levels of two genes are very stable (Figure 3B)
or are highly dispersed (i.e. do not overlap) (Figure 3C), their
expression ranking order will also have a high probability of being
preserved. As presented in Figure 4, all three factors contributed
significantly to the ? t tgg0 observed for the three gene sets, but their
total contribution decreased on going from HK (96.4%) to MR
(85.7%) to TS (53.7%); in the case of the TS genes, factors other
than the three considered contributed almost as much (46.3%) to
their ? t tgg0 value of 0.41. Of the three factors, dispersion contributed
the most, followed by stableness, then co-expression, except in the
TS set in which co-expression contributed somewhat more than
stableness (17.9% vs. 12%). However, the stableness value showed
the largest difference between sets, in that its contribution to the
HK set (70%) was more than twice that to the MR set (33.8%) and
five times that to the TS set (12%). A joint contribution of two or
three factors was particularly marked for the HK genes, suggesting
that their expression profiles depended on multiple characteristics
to a greater degree than the other two gene sets.
Stratification by Kyoto Encyclopedia of Genes and
Genomes (KEGG) pathways
To investigate whether the preserved ranking order of gene
expression resulted from biological regulation, we mapped the
genes to KEGG pathways, in which genes performing related
functions as categorized by biological pathways are grouped [26].
Of the 198 human pathways annotated in the KEGG, we
removed 37 belonging to the disease class and 6 with no more than
2 genes, leaving 155 pathways for analysis. Two hundred and
seventy-four of the 388 HK genes were annotated to belong to
these pathways, and the vast majority (264, or 96.4%) of these was
distributed in 7 pathways; these included 47 genes that were also
found in one or more of the remaining 148 pathways. As shown in
Table 2, the 7 pathways were enriched in HK genes from the
manually curated set, the percentage of HK genes in each of the 7
pathways ranging from 34% to 100%, considerably higher than
would be expected from an unbiased sampling of all genes (p
values for these percentages were statistically significant, ranging
from 0.04 to 0.006; see Table 2). Furthermore, the 7 HK-enriched
pathways link molecular biology’s central processes, i.e. from gene
Figure 1. HK genes exhibited a significantly more preserved
expression ranking order than MR or TS genes. Frequency is the
percentage of tissue pairs with the indicated Kendall’s tau (ttt0). Each
distribution is a compilation of CNt~36
2
(3)), where each ttt0 was for a pair of tissues sampled from a 36-tissue
pool and the frequency of ttt0 at a particular value was computed on a
histogram using a bin width of 0.1. The average value (? t ttt0) of the ttt0
distribution for the HK, MR, and TS gene sets was 0.77, 0.59, and 0.41,
respectively, and the standard errors were all small, mostly less than
0.01.
doi:10.1371/journal.pone.0029314.g001
~630 Kendall’s tau (ttt0; Equation
Preserved HK Gene Expression Ranking
PLoS ONE | www.plosone.org2December 2011 | Volume 6 | Issue 12 | e29314

Page 3

transcription (Basal Transcription Factors and RNA Polymerase
II) to mRNA processing (Spliceosome) to protein translation
(Ribosome and Aminoacyl-tRNA Biosynthesis) and degradation
(Ubiquitin Mediated Proteolysis and Proteasome). In comparison,
of the 12,687 NHK genes (see Methods), only 176 were found in
the 7 HK-enriched pathways, while 3,602 were found in the
remaining 148 pathways. The enrichment of HK genes in the 7
pathways was not surprising, since KEGG, along with Reactome
[27], was used to identify genes with essential cellular functions
[23].
Table 2 shows that, except for the Proteasome pathway
(0.5960.00), the Kendall’s taus (? t tgg0) of expression ranking of
the HK genes within their respective pathway (0.6560.00 to
0.7560.00) were all significantly higher than that (0.5860.00) of
the 3,602 NHK genes in the other 148 pathways, again revealing a
statistically significant difference in the preservation of their
expression ranking. Furthermore, co-expression correlation (? r rgg0)
in these HK-enriched pathways was mostly moderate, about 0.2 or
0.3, in accordance with the observation made earlier that co-
expression was a smaller contributing factor than either stableness
or dispersion to expression ranking (Figure 4). The exception was
the Ribosome pathway, which exhibited a high ? r rgg0 (0.7060.00)
and, as a result, a high ? t tgg0 (0.7560.00). A high ? t tgg0 (0.7560.00)
was also obtained for HK genes sampled from different pathways
(one from each pathway), which can be attributed to the much
larger range of expression levels exhibited between different
pathways than within a single pathway (Figure S2).
Decomposing ? t tgg0 into the three contributing factors for the 7
HK-enriched pathways showed a distribution that was similar to
that observed for the 388 HK genes (Figure 4), in that both
dispersion and stableness contributed significantly (,70–80%),
while co-expression was a relatively minor factor, with only a
,20–30% contribution (Figure 5). The high contribution of co-
expression in the Ribosome pathway, resulting from a high co-
expression correlation, as mentioned above, and the comparatively
low contribution of stableness for the Basal Transcription Factors
pathway are notable departures that merit further investigation.
Figure 6 shows that the expression levels of HK genes were
generally higher than those of NHK genes, an observation also
made by others [28]. Stratification of the results into the 7 HK-
enriched pathways revealed that, of these HK genes, those in the
Ribosome pathway had the highest expression levels and those in
the Basal Transcription Factors pathway the lowest. This explains
the corresponding highest and lowest contribution of stableness in
these two pathways (Figure 5), since high expression levels can
withstand expression variation more than low expression levels.
During transcription, the transcription factors of the Basal
Transcription Factors pathway mediate the binding of RNA
polymerase II to trigger initiation of transcription. Given that
Figure 2. Ranking preservation of HK genes was not a random event. (A) Kendall’s tau computed for the manually partitioned HK and MR
genes (circles) compared to those generated from two different random distributions: the triangles show the results when the GSE2361 expression
data were randomly divided into two sets containing the same number of HK (Ng=388) and MR (Ng=11,953) genes, while the squares show the
results when expression levels were ignored and rankings were randomly created (see Methods). Each data point is a combination of the two ? t ttt0
computed for the two gene groups (Ng=388 and Ng=11,953) for a particular pair of tissues. (B) The expression of all the genes in any two human
tissues (heart and thymus are shown as an example) always has an elliptical shape, resulting in a substantially preserved ranking order with a non-
zero ? t ttt0 (.0.4) even for randomly grouped genes (triangles in (A)). In contrast, randomly assigned rankings would yield a ? t ttt0 value very close to zero
(squares in (A)). Gene expression levels were log2 transformed and were denoted by xgt.
doi:10.1371/journal.pone.0029314.g002
Table 1. Kendall’s tau computed for the HK, MR, and TS
genes, for four datasets.
DatasetData type Kendall’s tau (? ttt0)
HK0.7760.00
GSE2361 [24] MR0.5960.00
TS0.4160.00
HK 0.6960.00
GSE1133 [50]MR0.4760.00
TS0.3460.00
HK0.7960.01
GSE803 [51]MR0.6260.01
TS0.2660.00
HK0.7760.00
Human BodyMap 2.0 data [53] MR0.5960.00
TS0.4260.00
doi:10.1371/journal.pone.0029314.t001
Preserved HK Gene Expression Ranking
PLoS ONE | www.plosone.org3December 2011 | Volume 6 | Issue 12 | e29314

Page 4

initiation is a rapid step in transcription [29], and, as shown in
yeast, the transcription of some genes is not dependent on basal
transcription factors [30], genes in this pathway may not need to
be expressed at high levels. In contrast, the ribosome is necessary
for translating all protein-coding genes, and, because translation is
a time-consuming process and relies greatly on the cooperation of
multiple ribosomal proteins [31], high expression levels of genes in
this pathway would be expected. However, interestingly, the 79
HK genes in the Ribosome pathway (Table 2) exhibited different
levels of expression in tissues of different embryonic origin
(Figure 7), their mean expression levels in tissues with an ectoderm,
mesoderm, or endoderm origin being 10.2460.16, 10.8060.15,
and 10.9060.08, respectively. Most of the ectoderm-derived
tissues constitute the mature nerve system, in which fewer gene
products are expressed continuously [32] and transcription of the
ribosomal genes need not be as active as in, say, bone marrow, a
mesoderm-derived tissue that is a factory generating blood cells,
thus requiring continual activation of the ribosomal genes, or the
endoderm-derived saliva gland, in which the production of
salivary amylase requires high utilization of ribosomes. Conse-
quently, the expression levels of these HK genes in the Ribosome
pathway are highly correlated across tissues of different develop-
mental origins, leading to the observed high co-expression
correlation (0.7060.00) (Table 2).
Finally, we can consider what could be responsible for the low
? t tgg0 (0.5960.00) for the Proteasome pathway (Table 2), despite a
typical contribution distribution from the three factors (Figure 5)
and typical expression levels of HK genes (Figure 6) in this
pathway. Further stratification of the Kendall’s tau results by the
subcomplexes of the proteasome (Figure 8) indicated that the 7
HK genes producing the b subunit of the 20S proteolytic core
particle were the culprit. It appears that the expression of not all
the proteins in the subcomplexes of proteasome is coherently
regulated [33]. For example, while the 7 gene products of the a
subunit are needed in equal amounts to form heptamer rings [34],
the 7 gene products of the b subunit cannot form rings by
themselves [35]. In fact, the 7 b-subunit proteins tend to remain in
the monomer state and often exhibit TS expression [36]; as a
result, there was a reduced preservation of the ranking order of
their expression across tissues, which, in turn, resulted in a
decrease in the ? t tgg0 for the Proteasome pathway (Table 2 and
Figure 8).
Discussion
Activation or modulation of regulatory events triggered by
different stimuli, such as hormones, transcription factors, or other
environmental changes, results in different levels of expression of
the same gene in different cells or tissues and, therefore, one would
not necessarily expect tissue-wide gene expression profiles to
exhibit a preserved ranking order. However, in this study, we
showed that the ranking order for the expression levels of the HK
genes was significantly more preserved than that of the NHK
genes in human tissues (Figure 1; Table 1). This observation was
substantiated by using data obtained from different gene
expression technologies (oligonucleotide microarray and short-
read RNA sequencing; Table 1), as well as an alternative set of HK
genes [37] (Figure S3).
Figure 3. Schematic illustration of the three factors that contribute to preserve gene expression rankings. The three factors represent
three different types of expression pattern of a pair of genes in different tissues: (A) Co-expression (rgg9; Equation (5)), (B) stableness (Sgg9; Equation (7)),
and (C) dispersion (Dgg9;Equation (8)).
doi:10.1371/journal.pone.0029314.g003
Figure 4. The contributions to the ranking preservation of the HK, MR, and TS genes. The three contributing factors were dispersion (solid
line), stableness (dashed line), and co-expression (dotted line). For each factor, its contribution (in parenthesis) was the sum of the individual
contributions calculated using tensor product decomposing equations, such as Equation (11)–(13). The contribution of factors other than these three
is given in the black box in the upper right corner of each panel.
doi:10.1371/journal.pone.0029314.g004
Preserved HK Gene Expression Ranking
PLoS ONE | www.plosone.org4December 2011 | Volume 6 | Issue 12 | e29314

Page 5

Tensor product analysis further revealed that dispersion, which
results from minimal overlaps due to, for example, a wide range of
expression levels, and stableness, a previously recognized hallmark
of HK genes, were two major factors underlying the observed
expression ranking preservation, whereas, perhaps unexpectedly,
co-expression made a relatively minor contribution (Figures 3 and
4). However, closer examination showed that, in certain pathways,
such as the Ribosome pathway (Table 2), or in the highly
collaborative expression of the proteins in a subcomplex of a
protein complex, such as the a subunit subcomplex (see Figure 8)
of the proteasome (? r rgg0~0:66+0:03, data not shown), co-
expression was indeed prominent. Furthermore, to a large extent,
the HK genes exhibited preserved tissue-wide expression rankings
with contributions from all three factors, especially dispersion and
stableness (Figure 4). Together, these results suggest that,
compared to NHK genes, HK gene expression is regulated more
consistently from tissue to tissue and that different mechanisms
may be involved in the regulation of different functional groups of
HK genes.
It has been shown in various studies that, compared to NHK
genes, HK genes tend to have a shorter coding sequence [37],
fewer exons and shorter exons [37], and a higher GC content [38].
However, while these observations are statistically significant as a
whole, these properties are poor measures for distinguishing
between HK and NHK genes due to the large overlap between the
two in terms of these properties (Figure S4). Moreover, it is difficult
to reconcile how these static properties of genes could confer the
differences in expression level from one tissue to another, let alone
the ranking preservation. In this study, we have uncovered a new
property of preserved expression ranking in different tissues, in
which grouped HK genes and NHK genes show a significant
difference. However, some overlaps between HK and MR genes
in their expression ranking were also evident (Figure 1), suggesting
that, to identify a gene as an HK gene by computational methods,
a composite index comprising multiple properties is probably
required. Regardless of whether or not the genes considered are
HK genes, the use of rank-invariant genes extracted from multiple
experiments as normalization references could reduce systemic
distortions in microarray data more than conventional treatments
[39,40]. While it may not be practical to use a global rank-
invariant set of gene transcripts, which numbers in thousands [40],
as references for real-time PCR experiments, a few of highly rank-
invariant genes, those within the same pathway (e.g. Ribosome) in
particular, may prove to improve the current protocol of such
experiments, but validation of this proposition requires experi-
mental investigations. Furthermore, the tensor structure of human
HK gene expressions uncovered in this work (Figure 4) can be
useful features for machine learning techniques to develop a
classification scheme for discovering novel HK genes (work in
progress).
The differences in HK gene expression in different tissues may
arise for a number of reasons. One is that the specific function of
tissues may dictate the level at which a HK gene needs to be
expressed. For example, and as shown in Figure 7, ribosomal
genes are expressed at a higher level in bone marrow cells,
Table 2. Kendall’s tau (? t ttt0) and co-expression (? r rgg0) computed for the HK genes in KEGG pathways.
KEGG pathway (number of genes; % in the 388 HK gene set)a
Expression rankingb? ttt0 (mean±SE)d
Co-expressionc? rgg0 (mean±SE)d
Basal Transcription Factors (33; 69.7%)0.6960.000.2260.01
RNA Polymerase II (11; 100.0%)0.7060.010.2060.01
Spliceosome (113; 46.9%)0.6760.00 0.3060.01
Ribosome (84; 94.1%)0.7560.000.7060.00
Aminoacyl-tRNA Biosynthesis (30; 63.3%)0.6660.000.2560.01
Ubiquitin-Mediated Proteolysis (117; 34.2%)0.6560.00 0.2360.01
Proteasome (42; 92.9%) 0.5960.00 0.3360.01
aBased on the distribution of the percentage of HK genes for each of the 155 KEGG pathways, the p value for a 34% was calculated to be 0.04, and 0.006 for 100%.
bWith the exception of the Proteasome pathway, these values are statistically significant (p value,10220by the paired two-sample Student’s t test) compared to the
value of 0.5860.00 obtained for the NHK genes mapped to the remaining 148 KEGG pathways. The p value for the Proteasome pathway was 7.6761024.
cAll these values are statistically significant (p value,10220by Student’s t test) compared to the alternative hypothesis of no correlation.
dSE: standard error.
doi:10.1371/journal.pone.0029314.t002
Figure 5. The contributions to the ranking preservation of the
HK genes in seven HK-enriched KEGG pathways. Shown is the
percentage contribution of the three factors, stableness, co-expression,
and dispersion to the expression ranking (Kendall’s tau, ? t tgg0) computed
for the HK genes found in each of the seven HK-enriched KEGG
pathways. The black bars are the contributions of other unknown
factors.
doi:10.1371/journal.pone.0029314.g005
Preserved HK Gene Expression Ranking
PLoS ONE | www.plosone.org5December 2011 | Volume 6 | Issue 12 | e29314