By giving greater weight to genes whose products are nearest to the terminal gene products of a pathway, the HER score gives more weight to genes specific to a particular pathway. This is illustrated in Figure 1, which uses the concept of terminal gene products. They are gene products like either receptors that initiate the pathway activity or transcription factors that are made to initiate transcription as a result of the pathway activity (see Materials and methods for a more mathematical definition). The genes involved in each of the signaling pathways in the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway database 44 were evaluated for the position of their gene products with respect to the terminal gene products and the total number of signaling pathways that these genes are involved in. It is clear from Figure 1 that genes associated with products that are closer to the terminal gene products are more pathway-specific.

To illustrate the utility of the DR as a scoring method, we consider the linkage between the full set of pathways in KEGG 44; that is, the pathways themselves can be viewed to be part of a higher level network, the nodes of which are pathways while the edges indicate the transfer of signal or material between pathways (Figure S1 in Additional data file 2). For example, the MAPK signaling pathway and the p53 signaling pathway can be considered to be linked. It seems reasonable to expect that after perturbation of the system, the affected pathways that are linked are more likely to respond similarly. We test this intuition using different microarray data (from the Gene Expression Omnibus (GEO) database 45 in a statistical test on the above network of pathways. The details are provided in the Materials and methods section. The P-values for the eight comparisons (estimated using 1,000 random networks) are given in Table 1. Significant P-values across the comparisons support our use of the DR as a reasonable score for differentiating between pathways.

Simulated data were generated from two pathway networks having different patterns of correlation between the various genes in the pathway, with each network having genes in a pool of genes representing a biological system. The pair of networks and the correlation patterns of genes in the pathway, denoted by pattern numbers, are listed in Table 2. Patterns 1, 2, 3 and 4 have non-zero correlation between a subset of genes in the system. All genes in pattern 5 are assumed to be independent of each other. Patterns 1 and 3 are biased to the scoring rules proposed here whereas patterns 2 and 4 are not. The treatments had the effect of increasing (as given in the variable, pert) the expressions of certain genes in the system.

Table 3 gives estimates of the type 1 errors of the five methods, at the 0.01 and 0.05 significance levels, for patterns 1 and 5. Table 4 gives estimates of the power of the SEPEA_NT1, GSEA and SEPEA_NT2 methods at 0.01 and 0.05 significance levels, for a pert value of 1.2 and for patterns 1-4. The empirical sizes of the methods maxmean and SEPEA_NT3 do not match their nominal sizes. So the results are provided at empirical sizes of 0.07 and 0.05 (corresponding to a nominal size of 0.001 for both cases).

Only patterns 1 and 5 were used to analyze the type 1 error behavior because they represented the two scenarios (presence or absence of gene-gene correlations) where pathway enrichment methods have been shown to have different behaviors 410. Because of the presence of correlations in the data, SEPEA_NT3 gives an incorrect type 1 error value for pattern 1 (Table 3). As has been stated previously, in spite of this incorrect behavior, there are situations (like those in which the only information available for each gene is a summary statistic representing the effect of the treatment) where methods like SEPEA_NT3 need to be used in order to create relevant hypotheses regarding affected processes due to the treatment. SEPEA_NT1, SEPEA_NT2 and GSEA do maintain the right type 1 error behavior in both the presence and absence of gene-gene correlations. In the presence of gene-gene correlations, the maxmean method 10 also does not maintain the appropriate type 1 error behavior. As expected, the power estimates of all three SEPEA methods for patterns 1 and 3 were significantly higher (P < 0.05, two-sample test of proportions) than those for patterns 2 and 4, respectively. The power estimates for patterns 1 and 3 using SEPEA_NT1 were higher than those for GSEA, demonstrating improvement in the ability to detect these biologically relevant patterns. For the other two 'not-so-relevant' patterns (2 and 4), SEPEA_NT1 was not always more powerful than the GSEA method. This loss of power can again be explained by the bias of SEPEA to detect conditions favored by the scoring rules. For example, the power estimates of SEPEA_NT1 were also higher than those for GSEA 35 for pattern 2 whereas this was not the case for pattern 4. At an empirical size of 0.07, maxmean does not appear to be competitive with the other methods. SEPEA_NT1 also provides a more powerful method than GSEA on pattern 1 across a range of perturbation levels and signal to noise levels (Tables S3 and S4 in Additional data file 1). In addition, power results for four other correlation patterns are presented in Table S2 in Additional data file 1.

The study by Raponi et al. 46 analyzes gene expression data taken from 130 lung cancer patients in different stages of the disease. They also provide survival times for each patient. The data are divided into two groups of 85 patients (training set) and 45 patients (test set). This was done such that the proportion of patients in each stage was approximately the same for the two groups. Using these data, the Cox proportional hazards statistic is computed for each gene on the microarray (indicating how predictive it is of the survival time of a patient). The next logical step is then an attempt to find what biochemical pathways are predictive of survival. All of the human KEGG 44 pathways are used in this analysis. The methods used were SEPEA_NT1, GSEA and maxmean. Also, to estimate the value of including information on the network structure, SEPEA_NT1 was applied to the data assuming that all the genes in the pathway are given equal weight and the DR score is zero. This analysis is denoted by SEPEA_NT1*. The goal of our analysis is to evaluate consistency in choosing 'significant' pathways found using the training set versus the test set. Curves for sensitivity versus '1 - specificity' and positive predictive value versus negative predictive value are obtained by using different cut-offs for the log of the P-values obtained using each method; the results are shown in Figure 2. The sensitivity, specificity, positive predictive and negative predictive values for SEPEA analyses have better ranges than those for GSEA and maxmean. For a significant portion of the ranges of sensitivity and specificity for GSEA and maxmean, the SEPEA analyses provide higher sensitivity for a given level of false positives (a point on the '1 - specificity' axis). The same can be said about the portion of the ranges of the positive and negative predictive values of maxmean dominated by the SEPEA analyses. From the curves for SEPEA_NT1 and SEPEA_NT1*, we also observe the benefit of incorporating pathway network information. An updated Figure 2 that also includes results from SEPEA_NT2 and SEPEA_NT3 is provided as Figure S2 in Additional data file 3.

Enriched KEGG pathways using SEPEA_NT2 and SEPEA_NT2* (which is essentially the SEPEA_NT2 analysis but does not make use of the network information of the pathways and is identical to the analysis of the Q2 test in 9) methods were determined for a microarray dataset (see Materials and methods section) examining the consequences of inhibition of Sonic hedgehog (SHH) signaling by cyclopamine treatment of developing Xenopus laevis (Tables 5 and 6). Based on the specificity of cyclopamine to inhibit the SHH pathway, we expected to see the SHH signaling pathway significantly enriched; however, the P-value for this pathway was not significant using either method (SEPEA_NT2 and SEPEA_NT2*). This may be due to the time point at which gene expression was evaluated, which was optimized to evaluate downstream effectors of SHH pathway inhibition. Alternatively, this result may also reflect the limitation of the method when using only gene expression datasets, as several components of the SHH pathway, including Hedgehog (Hh) and Patched (PTCH), are known to be regulated at the protein level. Finally, when results obtained using SEPEA_NT2 versus SEPEA_NT2* are examined in the context of pathways linked to the SHH pathway (Figure S1 in Additional data file 2), we see that only the MAPK and Proteasome pathways are reachable from the SHH pathway by two and three edges, respectively, suggesting that results from SEPEA_NT2 may be more consistent with targets downstream of the SHH pathway. None of the other pathways listed in Tables 5 and 6 were reachable along the network of pathways (Figure S1 in Additional data file 2) from the SHH pathway. In fact, recent evidence suggests that SHH promotion of proliferation and differentiation in muscle 47 and gastric mucosal cells 48 is through transcription-independent activation of the MAPK/ERK pathway. This analysis suggests benefits of using pathway network information. Additional results from analysis of these data with SEPEA_NT1, SEPEA_NT3, GSEA and maxmean are provided in Additional data file 4.

Genes associated with breast cancer were downloaded from the Online Inheritance in Man (OMIM) database 49. This group of genes was pruned to include only those genes that participate in a pathway in the KEGG pathway database 44. The list of genes used is provided in Table S5 in Additional data file 1. The SEPEA analysis was used to test whether there is an overabundance of 'important' (as defined by the scoring rules) breast cancer genes in pathways relative to the remaining set of genes that participate in some pathway in the KEGG pathway database 44. Using these data, SEPEA_NT3 and SEPEA_NT3* (which is essentially the SEPEA_NT3 analysis but does not make use of the network information of the pathways and is very similar to those used in 79111213141516171819202122) was used to find the enriched human pathways associated; the results are given in Table 7. Several of the pathways known to be important for breast cancer initiation and progression are significant using either method, such as the ErbB, p53, and apoptosis pathways. In contrast, the adherens junction, regulation of actin cytoskeleton, cell adhesion molecules, and focal adhesion pathways are significant using SEPEA_NT3, but are not considered significant using the SEPEA_NT3* method (P ≤ 0.05). These pathways, in particular the focal and cell adhesion pathways, all deal with cell to cell communication and are thought to be key modulators of progression and invasion of malignant phenotypic characteristics 50. In fact, several novel cancer chemotherapy drugs are being designed to specifically act on the focal adhesion pathway and many standard chemotherapy drugs modulate this pathway in conjunction with their primary mode of action 51. So this analysis again suggests gains in the pathway enrichment analysis when network details of pathways are incorporated in the analysis.

Assume there are m genes (identified by indices in the set G = {1, 2,..., m}) in the system and n array measurements (n_c control and n_t treated, n_c + n_t = n) per gene. We will analyze one particular pathway made up of a subset m_P of the m genes in the system. Without loss of generality, assume that these genes correspond to the first m_P gene indices in G. The genes in this pathway are part of an underlying network of their gene products. On the basis of this network, gene i of the pathway is assigned a weight w_i and a gene pair (i and j) is assigned two weights d_ij (denoting a measure of the distance between these two genes on the network) and e_ij (which is equal to 1 for a non-zero value of d_ij). Each of the m genes is also assigned a value, t_{stat, k} for gene k capturing the treatment effect on it as found in the observed data. This value obtained under the different null distributions (as defined in the next section) is denoted by T_{stat, i}. The two scores, from the Heavy Ends Rule and the Distance Rule are denoted by HER and DR, respectively. They are a function of t_{stat, k}. HER_obs and DR_obs denote those obtained from the observed experimental data while HER_rand and DR_rand those obtained from the different null distributions.

Null hypotheses for the three statistical tests performed are given below and share similarities with those stated in 6.

Network test 1 (NT1): T_{stat, i}, i = 1, 2,...m are identically distributed (and possibly dependent) with common distribution, F₀ corresponding to the lack of association with the treatment, for each gene.

Network test 2 (NT2): T_{stat, i}, i = 1, 2,...m_p (only genes in the pathway) are identically distributed (and possibly dependent) with common distribution, F₀ corresponding to the lack of association with the treatment, for each gene.

Network test 3 (NT3): T_{stat, i}, i = 1, 2,...m are independent and identically distributed with a common distribution, F (which can take any form).

In all three hypotheses, HER_obs and DR_obs are each drawn from the distribution of HER_rand and DR_rand, respectively.

For each gene we define by a pair of values (, ) corresponding to the association with the treatment in the context of the observed data. The association of any given gene with treatment is given in terms of the square of the two-sample t-statistic (similar to what has been done in 62535) and also shares similarities with the maxmean statistic defined in 10. Mathematically:

where , are the sample mean gene expression for gene g_i of the control and treated data, respectively, , are the associated standard deviations, I_NT1 is equal to 1 when the NT1 test is being used and is equal to zero otherwise, denotes the position of gene i in the sorted (in descending order) list of max(t_{stat, k}, 0) over all the m genes, and, similarly, denotes the position of gene i in the sorted (in ascending order) list of min(t_{stat, k}, 0). a and b are parameters chosen empirically in order to control for the selection of the pathway with the most significant genes (relative to the other genes in the system). The first terms in the products on the right-hand side of Equation 2 will be called importance factors for a gene. These are values between 0 and 1. The functions 'mean' and 'var' refer to the standard definitions of mean and variance. The term CF denotes a (competitive) factor that is a measure of difference in the mean of differential expression of the genes in the pathway and that of the other genes in the system. Higher CF values indicate higher individual association values for genes in the pathway relative to the other genes and vice versa. Therefore, for similar values for changes in gene expression (t_{stat, i} s) the power to detect treatment effect decreases as the CF factor decreases (or as more genes in the system are affected as a result of the treatment). For high values of the CF factor, parameter a controls the (decreasing) importance of genes along the sorted list. The parameter b provides a much steeper decrease in the importance of genes down the sorted list for small values of the CF factor.

Here, t_{stat, i} is the standard two sample t-statistic. In some instances, the only information of the association of a gene with a treated condition may be just a summary statistic. For example, there are a set of known gene polymorphisms associated with breast cancer; in trying to identify pathways relevant for breast cancer, these genes would then be arbitrarily assigned a t_{stat, i} equal to 1 while the other genes would be given values of 0. Note that in these situations, n, the number of array measurements per gene, is zero.

The score for linking the observed expression data to a given pathway has two components. The first component is called the Heavy Ends Rule score HER_obs and will have a high value when a combination of the more 'important' genes (those associated with gene products close to a terminal of a pathway) is significantly associated with the treated condition. The second component called the Distance Rule score DR_obs has a high value when the genes that are significantly associated with the treated condition have their gene products located close together. It is in fact the reciprocal of the weighted average distance between the genes in the network. The weights w_i, d_ij and e_ij are defined in a subsequent section. Each score is defined as the maximum of individual expressions dependent either only on the genes whose expression increased due to the treatment or on the genes whose expression decreased as a result of the treatment. This should make it more robust to detect changes in both scale and location as discussed in 10. The two scores are defined as:

For the DR score computation, 0/0 is defined to be equal to zero. The scores obtained under the null distributions are denoted by HER_rand and DR_rand and are defined as in Equation 4 with t_i replaced by T_i.

For each of the three hypotheses (NT1, NT2 or NT3) the test statistic is defined as:

where mean(HER) and std(HER) refer to the mean and standard deviation of the HER score for the given test and mean(DR) and std(DR) are those for the DR score.

For the NT1 and NT2 tests, multiple random samples of arrays are taken from the common set of treated and control data (without replacement) and randomly assigned to control or treated groups. For each random sample, the T_{stat, i}s are calculated and then HER_rand and DR_rand are computed. The NT1 test requires T_{stat, i} to be computed for all the m genes while the NT2 test requires computation for just the m_P genes that are part of the pathway. For the NT3 test, multiple random samples of m_P T_{stat, i} s are drawn from the global set of m observed t_{stat, i}.

The estimate of the P-value for each of the tests is computed as:

where I(S_i ≥ S_obs) is an indicator function that equals 1 when the i^th randomly estimated test statistic value, S_i, equals or exceeds the observed value and 0 otherwise. The estimation procedure used for the special case when the data are in the form of a list of differentially expressed genes or a list of genes associated with a disease is provided in Additional data file 1.

The way the significance computations are performed, tests NT1 and NT3 could be viewed as belonging to the class of 'competitive' hypotheses (as elaborated in the Background section) while NT2 could be viewed as a 'self-contained' hypothesis.

The method when applied to each of the three null hypotheses NT1, NT2 and NT3 is denoted by SEPEA_NT1, SEPEA_NT2 and SEPEA_NT3, respectively.

Data were simulated from two genetic systems (Linear (L) and ErbbSignaling (E)) of 500 genes ( and ). Each system had two subnetworks of interest and each subnetwork was assumed to have no interactions with the other subnetwork. The Linear network had a set of 30 genes that were connected in a linear fashion (Figure 3a). A set of 87 genes in the ErbbSignaling network interacted in the same manner as described by the Erbb signaling pathway in the KEGG pathway database 44 (Figure 3b). Pathway enrichment analysis was performed on these two subnetworks.

Each set Λ and H had a subset of genes (with indices ), , whose expressions were perfectly correlated with each other (Σ^L had n_corr = 0 or 9 genes and Σ^E had n_corr = 7 genes). The gene expressions in the complement of each of the sets Σ^L and Σ^E, (Σ^L)^c and (Σ^E)^c, were assumed to be independent of each other even though some of them could be assumed to be known to have gene products that interact with gene products of genes in Σ^L and Σ^E. This could be justified by the fact that the interaction was not at the gene expression level and involved changes in the phosphorylation/binding states of the protein, for example. Let denote the set of gene indices associated with the proteins circled in Figure 3b, ordered from left to right. The random variable defining the gene expression of gene g_n is denoted by X_n. Let N(μ, σ) represent the normal probability distribution with mean μ and standard deviation σ. Then data for all the 500 genes in each of the two systems were generated for one experiment under control conditions in the following manner:

Let Φ (Φ^L and Φ^E) denote the set of genes that are direct targets of the treatment. The total number of genes in the system affected by the treatment (that includes the set Φ) was chosen to be 50 and 10 for the Linear and ErbbSignaling networks, respectively. The effect of the treatment was to increase the mean of the expressions of the direct targets by a factor pert, μ' = pert·μ. Results from the assignment pert = 1.2 are discussed here while those resulting from other assignments are discussed in Table S3 in Additional data file 1. Let U^L and U^E denote a uniformly random selection of n_corr genes from the sets Λ and H, respectively, let V_n^L and V_n^E denote sets of n genes drawn from the complements of the sets Λ and H, respectively, and let Ø denote the empty set. The details of the different correlation patterns considered here are given in Table 1. Patterns 1 and 3 were the correlation patterns that were favored by the scoring rules described in this paper.

All methods in this paper were coded using the Java programming language. For each combination of correlation pattern and pert assignment, 1,000 independent experiments were simulated. Each experiment involved the generation of n_c = 5 control samples and n_t = 5 treatment samples. For the randomization tests of each method, 1,000 randomizations were chosen. The performance measures chosen were the number of experiments out of the 1,000 performed that resulted in P-values for the test (Equation 6) below different chosen significance levels. The methods evaluated were GSEA 35, maxmean 10, SEPEA_NT1, SEPEA_NT2 and SEPEA_NT3. For the SEPEA_NT1 method, the parameters a and b in Equation 2 were empirically set to equal 2 and 5, respectively. Parameter a = 2 provides a quadratic decrease in the importance of genes along the sorted list for high values of the CF factor (when the mean changes in expression of the genes in the pathway are higher than that of the rest of the genes in the system). In the situation of low values of the CF factor, the value b = 5 was chosen such that the top 20% of genes in the sorted list approximately receive importance in the interval (0.2, 1) while the remaining genes receive weights in the interval (0, 0.2). Results from GSEA 35, maxmean 10 and SEPEA_NT1 are comparable because all test a similar null hypothesis. The main difference between these methods is that while GSEA and maxmean are blind to the structure of the biochemical pathway, SEPEA-NT1 is not.

The pathway network is represented by a set of nodes/gene products and set of edges between these nodes. The nodes represent gene products such as individual proteins or protein complexes. There is an edge from node/protein u to node/protein v if u transfers the signal it received immediately to v (either in the form of increasing the transcription of genes associated with v, changing the phosphorylation state of v, causing disassociation of v from a complex that it is part of) in the case of signaling pathways or that u and v catalyze two successive reactions in the case of metabolic pathways.

Let denote the set of P nodes of the network and denote the set of N genes associated with the nodes. The number of edges entering node v_i is defined as its in-degree and the number of edges leaving v_i is defined as its out-degree. We define a node to be a terminal node if either its in-degree or out-degree is zero.

Assume that each edge represents a unit distance between the two nodes that it connects. So if the shortest route between two nodes is via two edges in the pathway network, then the two nodes are said to be 2 units of distance apart. Note the phrase 'distance between a pair of nodes' is used to imply 'shortest distance between this pair', considering that there may be more than one path connecting the two nodes in the pathway network. Let δ_j denote the shortest distance of node v_j to a terminal node of the pathway. Let G(v_i, g_a) denote the indicator function, which is equal to 1 when gene g_a is associated with node v_i and is equal to 0 otherwise. The number of genes associated with node v_i is denoted by N_i. Let s_ij denote the distance from node v_i to node v_j in the network. s_ij is assigned a value of 0 either when i = j or when node v_j is unreachable from node v_i. Define the positive indicator function, I⁺(x), which is equal to 1 when x is positive and equal to 0 otherwise.

The weights for gene g_a, w_a, and gene pair (g_a, g_b), d_ab and e_ab, are given by:

The weight w_a is defined such that genes associated with nodes closer to the terminal nodes have higher weights than those that are further away. The choice of a linear function to capture the intuition behind the HER is arbitrary and other functions will be experimented with as part of future work. The non-zero weights d_ab for genes a and b are smaller if they are associated with gene products that are closer together in the pathway network than for pairs of genes whose gene products are further away.

Let the total number of pathways (nodes) in the network in Figure S1 in Additional data file 2 be denoted by N_p. Denote the distance between pathways i and j on this pathway network by . Define to be equal to zero if pathway j is not reachable from pathway i. Also define variable , which is equal to 1 for all non-zero values of the corresponding and 0 otherwise. Perturbations to one pathway are transferred across the edges of the network to multiple pathways. Using human microarray data randomly chosen from the GEO database 45, we considered eight comparisons between two conditions (Table 1). For each comparison, the DR score was computed (Equation 4) for every human pathway on the network of pathways described above. In order to make the comparison possible across all the pathways, the DR scores obtained above using experimental data were normalized with DR scores obtained by setting the T_{stat, i} values for all the genes equal to 1. Let the normalized DR score for pathway i be denoted by . A meta score can now be defined on the pathway network as follows:

Higher values of meta_DR would indicate that pathways with higher values of the normalized DR scores are closer to each other. The significance of the obtained meta_DR scores are tested using random networks generated by the Markov-chain switching algorithm 52. The properties of these random networks are that they have the same number of nodes and edges as the original pathway network and the degree sequence among all the nodes is also maintained. These networks differ, however, from the original network due to a number of random edge swaps across the network.

Cyclopamine powder (11-deoxyjervine; Toronto Research Chemicals Inc., North York, Ontario, Canada) was dissolved in 100% ethanol to a concentration of 5 mg/ml and this stock solution was stored at -20°C. A similar volume of 100% ethanol was stored at -20°C for use in vehicle control exposures. Approximately 200 tadpoles from each of two clutches (designated 'clutch A' and 'clutch B') of the species Xenopus laevis were obtained from Nasco Biology (Fort Atkinson, WI, USA) for a total of approximately 400 tadpoles. Animals were raised at an air temperature of 25 ± 1°C in tanks of 9 liters of tap water treated with Stress Coat (Aquarium Pharmaceuticals, Chalfont, PA, USA) and aged 1 day. Each day for three consecutive days, as animals reached stage 52 53, the population of stage 52 individuals from each clutch was removed from the clutch tanks and divided in half indiscriminately, resulting in four exposure groups per day: a control group for clutch A; an experimental group for clutch A; a control group for clutch B; and an experimental group for clutch B. Each exposure tank had between 10 and 20 individuals, with 150 ml treated water per individual. After sorting into exposure tanks, 30 μl per animal of 5 mg/ml cyclopamine solution was added to all experimental tanks, and 30 μl per animal of 100% ethanol was added to each control tank. After 24 hours of exposure, animals were sacrificed by over-anesthesia with MS222, dried on a paper towel, then put into vials of RNAlater (Ambion, Austin, TX, USA). Vials were kept at 4°C overnight, then moved to -20°C for storage. Both hindlimb buds were dissected off each animal at the base of the limb using surgical scissors, placed in fresh vials of RNAlater, and returned to -20°C for continued storage. RNA extractions were performed using the RNeasy Mini Kit and optional RNase-Free DNase Set (QIAGEN, Valencia, CA, USA), with the following notes: limbs were put into a 1.5 ml microcentrifuge tube, residual RNAlater was pipetted off, and limbs were crushed with a homegenizer in 200 μl buffer RLT, then 300 μl more buffer RLT was added; and elution was carried out with two washes of 50 μl RNase-free water. Extracted total RNA was stored at -80°C and transferred to the WM Keck Foundation Biotechnology Resource Center, Affymetrix Resource Center (Yale University, New Haven, CT), where they were again run through DNase treatment. Four control-experimental pairs of samples were chosen, from a total of 12 pairs, based on quantity and quality of RNA as determined by analysis on an Agilent 2100 Bioanalyzer RNA Nano chip (Agilent Technologies Inc., Santa Clara, CA, USA). Samples in each pairwise comparison were extracted from the same number of limbs, were from the same clutch, were exposed to cyclopamine solution or ethanol on the same day, and their total RNA was extracted in the same batch of extractions. The eight chosen samples were each hybridized to an Affymetrix^® GeneChip^® Xenopus laevis Genome Array (Affymetrix, Santa Clara, CA, USA) using 3 μg total RNA. Data have been deposited in the National Center for Biotechnology Information, NCBI GEO with series record ID [GEO:GSE8293].