Using Stochastic Causal Trees to Augment Bayesian Networks for Modeling eQTL Datasets
- Kyle C Chipman^{1}Email author and
- Ambuj K Singh^{1, 2}
DOI: 10.1186/1471-2105-12-7
© Chipman and Singh; licensee BioMed Central Ltd. 2011
Received: 8 April 2010
Accepted: 6 January 2011
Published: 6 January 2011
Abstract
Background
The combination of genotypic and genome-wide expression data arising from segregating populations offers an unprecedented opportunity to model and dissect complex phenotypes. The immense potential offered by these data derives from the fact that genotypic variation is the sole source of perturbation and can therefore be used to reconcile changes in gene expression programs with the parental genotypes. To date, several methodologies have been developed for modeling eQTL data. These methods generally leverage genotypic data to resolve causal relationships among gene pairs implicated as associates in the expression data. In particular, leading studies have augmented Bayesian networks with genotypic data, providing a powerful framework for learning and modeling causal relationships. While these initial efforts have provided promising results, one major drawback associated with these methods is that they are generally limited to resolving causal orderings for transcripts most proximal to the genomic loci. In this manuscript, we present a probabilistic method capable of learning the causal relationships between transcripts at all levels in the network. We use the information provided by our method as a prior for Bayesian network structure learning, resulting in enhanced performance for gene network reconstruction.
Results
Using established protocols to synthesize eQTL networks and corresponding data, we show that our method achieves improved performance over existing leading methods. For the goal of gene network reconstruction, our method achieves improvements in recall ranging from 20% to 90% across a broad range of precision levels and for datasets of varying sample sizes. Additionally, we show that the learned networks can be utilized for expression quantitative trait loci mapping, resulting in upwards of 10-fold increases in recall over traditional univariate mapping.
Conclusions
Using the information from our method as a prior for Bayesian network structure learning yields large improvements in accuracy for the tasks of gene network reconstruction and expression quantitative trait loci mapping. In particular, our method is effective for establishing causal relationships between transcripts located both proximally and distally from genomic loci.
Background
In order to model and dissect the complexity underlying physiological processes, including diseases, developmental programs, and responses to pharmacological treatments, systematic approaches based on genome-wide data are imperative. Expression profiling technologies, such as microarray [1, 2] and RNA-Seq platforms [3], provide quantification of mRNA levels on a genome-wide scale, prompting computational methods aimed at learning a more holistic perspective of cellular processes. Parallel advancements in the area of genotypic profiling, including high-throughput sequencing and SNP detection, offer information complementary to that of expression data. These concurrent developments pave the way for genetical genomic studies, which provide the joint space of expression and genotypic data corresponding to offspring that arise from a segregating population [4]. To date, eQTL datasets have been published for several organisms [5–10], providing ample opportunity to develop novel computational methodologies. The tandem existence of expression and genotypic data is especially powerful in that it allows one to reconcile changes in expression programs in the context of the specific genetic combinations represented by the offspring. Since natural genetic variation is the sole source of perturbation, it is logical to view genomic loci as epicenters of phenotypic variation in eQTL-derived causal networks. Consequently, modeling eQTL datasets enables one to hypothesize on how genotypic variation results in phenotypic changes.
Already several studies have provided methodologies aimed at exploiting the genotypic component of eQTL data to improve causal modeling in gene networks [9, 11–16]. Bing et al. introduced methodology to build directed networks starting from a set of candidate cis-genes for each locus [14], establishing directed edges from candidate cis-genes to distally-located genes. This approach yields local regulatory models for individual loci, and the authors also present an innovative approach based on partial correlations to identify models where two regulators play complementary roles in controlling a common set of genes. The methodology of Bing et al. was later applied to an eQTL dataset representing Arabidopsis by Keurentjes et al., who also incorporated information regarding DNA sequence to improve the estimation of cis-genes [9]. While this application was successful in providing hypotheses regarding local regulatory models, it does not resolve causal orderings amongst distally located transcripts. Furthermore, modeling local regulatory programs with respect to individual loci leaves room for improvement in the sense that each of the respective models are disjoint. A worthwhile goal is to produce more holistic and systematic methodologies capable of modeling the complex interdependencies between multiple loci and transcripts. Indeed, it has been estimated that the genetic basis of many transcripts is extensively complex, with upwards of 50% of transcripts being linked to five or more loci [17]. The need for a comprehensive and systematic approach was addressed by Schadt and colleagues, who developed a novel method to augment Bayesian networks with probabilistic measures to direct causal orderings of gene pairs with respect to genomic loci [11–13, 18]. Their method, which is based on a conditional bivariate normal model, determines if two transcripts linked to a common locus are best modeled as causal or independent [12]. Ultimately, the information generated by their method is incorporated as a prior for Bayesian network structure learning [19]. This approach has yielded promising results when applied to yeast [13] and mouse [12], providing hypotheses regarding the architecture of eQTL networks. Furthermore, the authors published a study on synthetic networks to quantify the performance gains associated with their method [18] as compared to standard Bayesian network structure learning. While their method proved efficacious at resolving causal orientations between correlated transcripts in the context of a global network, the scope is generally limited to the upper echelons of the causal hierarchy, an attribute that stems from their reliance on using genomic loci as causal anchors. Ideally, one could commence at the genomic loci, learn the causal orderings of the most proximal transcripts, then advance down the causal hierarchy propagating the structural information gleaned from the upper levels of the network. The probabilistic method presented herein realizes this concept by stochastically reconstructing the causal hierarchy, which is subsequently incorporated as a prior into Bayesian network structure learning.
A distinct but related problem to gene network reconstruction (GNR) is expression quantitative trait loci (eQTL) mapping. While these two tasks have to date been addressed independently, they are likely to become more intertwined as eQTL-related computational methodologies advance. This corollary follows from the fact that accurately-modeled networks should inform on transcript-locus associations by virtue of the implied causal pathways. Traditional univariate methods, which involve an exhaustive search between all transcripts and loci, typically entail the use of linear regression, ANOVA, or the t-test [20], where the chosen statistic assesses the extent to which a trait (transcript level) is linked to a locus. While straightforward, this approach results in a considerable loss in statistical power due to the need to correct for multiple hypothesis testing. The issue of multiple testing is compounded when transcripts are tested for simultaneous linkage to two loci, where an exhaustive search across all loci results in complexity that is quadratic in the number of loci. Addressing this limitation, Storey et al. developed a stepwise mapping procedure [21] whereby testing linkage to a secondary locus involves conditioning on the primary locus, reducing the complexity to 2L - 1 tests, where L is the number of loci. This approach, which controls for confounding linkages and tests for epistasis, was later generalized for multiple interval mapping by Zou et al.[22]. While stepwise mapping techniques produce notable improvements over exhaustive searches, they do not take advantage of the fact that many transcripts sharing linkages to common loci exhibit strong correlation structure. Suitably, recent methods have focused on mapping multiple (correlated) traits simultaneously to genomic loci [23–26]. Pan et al.[25], Litvin et al.[24] and Lee et al.[23] developed methods based on penalized regression frameworks to discover modules that can be explained by common expression programs. Similarly, Zhang et al. developed a Bayesian partition method to learn modules of transcripts sharing linkage to common genomic loci [26]. These clustering based methods offer the distinct advantage of using fewer parameters while detecting eQTL linkages, however, they do not attempt to reconstruct the causal hierarchy and are therefore limited in the biological hypotheses that they offer.
Our goal is to reconstruct causal networks with high-fidelity at all levels of the network. Consequently, by improving the accuracy of the reconstructed network, we show that our method can provide biological hypotheses as well as enable greater accuracy in eQTL mapping.
Results
Network Parameters
Stronger Cor. Structure | Weaker Cor. Structure | |
---|---|---|
Mean CC | 0.68 | 0.55 |
90% Upper Interval | 0.88 | 0.76 |
90% Lower Interval | 0.45 | 0.31 |
Mean a_{ i } | 0.75 | 0.6 |
S.D. a_{ i } | 0.2 | 0.2 |
Mean k_{ i } | 0.5 | 0.5 |
S.D. k_{ i } | 0.1 | 0.1 |
Precision and recall of network edges
We note that the performance gains of LCMS-augmented structure learning over unaugmented structure learning reported by Zhu et al.[18] are recapitulated in our experiments, though the margin of improvement is slightly lower in our study. This is due to the fact that we explicitly model genomic loci as head nodes, which equates to the authors' logic whereby transcripts with cis-linkages are required to be head nodes. Consequently, the performances of the baseline "Basic" Bayesian networks are higher in our study as compared to Zhu et al.[18]. In experiments not published herein, we estimate that the performance margins between the Basic and LCMS-augmented methods are generally at least twice as large when this logic is not included into the unaugmented Bayesian network structure learning procedure, indicating that performance gains are realized by simply establishing the head nodes as such.
Precision and recall of transcript-locus associations
Next, we sought to assess the potential of using the learned networks for the purpose of expression quantitative trait loci (eQTL) mapping. In order to establish a transcript-locus linkage, for each head node (locus) in the Bayesian network, we run a depth-first search down the respective branches. All reachable transcripts from the source locus are associated with that locus. Starting with a set of sampled networks, we apply this procedure on each of the individual networks, each yielding a set of transcript-loci linkages. Precision-recall curves were generated from the totality of the individual networks (Methods section). We compare this approach to traditional univariate mapping utilizing a t-test. We note that while there are several recent clustering-based methodologies related to eQTL mapping [23–26], they do not attempt to reconstruct gene networks. Since GNR is our primary goal, we restrict our analysis to the three versions of Bayesian networks and traditional univariate mapping.
Robustness of network reconstruction
As outlined in the Methods section, we assess convergence and reconstruct consensus networks from two independent MCMC runs, each consisting of 150 million iterations. However, an additional test for robustness involves assessing the stability of edge frequencies across multiple MCMC runs. We follow the protocol outlined by Zhu et al.[18], whereby all edges above a predefined frequency threshold are extracted for five individual MCMC runs. Additional file 1 figures S 1a-S 1c provide the network reconstruction stabilities for datasets of weaker correlation structure and 200 samples, presented for each of the three methods: (a) unaugmented, (b) LCMS-augmented, and (c) SCT-augmented Bayesian networks. We adjusted the frequency threshold that determines the consensus network; the values are: 0.4, 0.6, and 0.8. For each method and for a given threshold, we first quantify the number of edges that occur in at least one of the five runs, denoted as T. T is the cardinality of the superset of consensus edges derived from the union of the individual sets corresponding to the five runs. We plot count(x)/T for values of x = {1, 2, 3, 4, 5}, where count(5) indicates that an edge is present in the consensus networks in each of the 5 MCMC runs. At the highest threshold of 0.8, each of the three methods showed robustness to where at least 80% of the total edges are present in the consensus networks for all 5 MCMC runs. These findings are consistent with those of Zhu et al.[18]. Finally, Additional file 1 figure S 1d depicts the number of edges that are recalled in all five MCMC runs at the three thresholds. The results suggest that our method offers relatively higher consistency across sampled networks.
Possible Explanations for Increased Performance
There are two likely reasons for the enhanced performance associated with our SCT method. The first and most obvious reason is the increased coverage associated with the SCT method. The second reason is attributable to better resolution of ordered triplets. For example, consider the hypothetical sequential triplet influenced by a single locus: L_{1} → t_{ i } → t_{ j } → t_{ k } . The LCMS procedure will generally establish the following relationships: t_{ i } → t_{ j } , t_{ i } → t_{ k } and t_{ j } → t_{ k } . One practical limitation of this approach is that it does not usually prefer the motif t_{ j } → t_{ k } over t_{ i } → t_{ k } . That is, both pairwise orderings are likely with respect to the locus L_{1}. In contrast, our method does not rely on a fixed anchor (genomic locus), and is capable of discriminating between potentially confounding motifs of this nature. This is due to the fact that, while both transitions of t_{ i } → t_{ j } and t_{ i } → t_{ k } are likely for our SCT method, the best scoring alignment corresponds to the true configuration, which will often be realized as the average configuration over multiple runs. In general, finding the optimal configuration will be self-perpetuating when various ordered triplets are interconnected, as the learning of causal relationships at lower rungs is clearly dependent on correct alignments in the upper levels.
Discussion
We presented methodology aimed at utilizing genotypic data for the task of gene network reconstruction on eQTL datasets. Our method is motivated by previous studies focused on the same goal, however, we are able to provide improvements in coverage and resolution. Furthermore, with enhanced network reconstruction accuracy, we show that sampling a set of networks is efficacious at eQTL mapping. Although we followed established protocols for simulating eQTL data, it's inevitable that the simulated data does not perfectly model natural eQTL data. For example, our model omits feedback loops, though such motifs are common in real gene networks. Furthermore, we clearly are unable to model cases where genetic variations are associated with amino acid substitutions without corresponding expression changes. Other situations that we are unable to model include post-translational modifications, such as protein-phosphorylations and other mechanisms affecting protein concentrations. It is worth noting that, since our model is generally more complicated than univariate mapping techniques, it stands to reason that univariate mapping might be less sensitive to discrepancies between the model used in our study and real eQTL networks.
Future work involves applying our methodology to datasets that incorporate macroscopic phenotypes, including medical conditions and responses to pharmacological treatments [27]. Reconciling variations in macroscopic phenotypes with genotypic and expression variations is an emerging problem that is poised to yield great insights into the molecular bases of complex phenotypes. Already methods exist that focus on predicting the outcome of pharmacological treatments in the yeast eQTL dataset [28, 29]. While the prediction of such outcomes are useful, we posit that accurate classification of an outcome is better viewed as a constraint within the general problem of learning a causal network. This formulation is more holistic and requires that learned networks agree with the classification of the macroscopic phenotype. Given that complex macroscopic phenotypes are often connected to several loci at relatively weak levels, our SCT method offers potential for this class of problems due to its ability to causally connect distal nodes in a network. Other possible applications include association studies, which carry an added degree of difficulty due to the genotypic heterogeneity associated with population-wide samples.
There are several possible ways in which our approach can be optimized. For example, we plan to investigate the use of iterative procedures, where information from prior runs is incorporated into subsequent runs to improve accuracy. With respect to the general area of Bayesian network structure learning, it would be interesting to consider integrating the causal ordering information from our method with other sources of prior biological information, such as protein-protein interactions or gene ontology (GO) annotation [30]. Such lines of inquiry should be feasible in light of the recent development of principled ways of incorporating prior information [31–33]. Finally, we note that our method utilizes parameters that are optimized a priori to structure learning. While this straightforward approach is able to provide appreciable performance improvements, further benefits might be realized by optimizing the parameterization of our method jointly with the beta parameter in the structure learning procedure; methodology addressing joint optimization problems in the context of Bayesian network structure learning was recently presented by Werhli et al.[33].
Conclusions
We developed a probabilistic method based on stochastic causal trees to learn the causal relationships between gene transcripts in genetical genomics studies. Incorporating the information from our method as a prior into Bayesian network structure learning increases the performance of network reconstruction and eQTL mapping.
Methods
Estimating Network Properties
The synthetic network consists of 2, 200 transcripts and 50 loci, connected by 2, 598 edges. These values were chosen based on analysis run on the yeast eQTL data published by Kruglyak and colleagues [17]. We estimated the number of transcripts that show significantly variable expression patterns in the yeast eQTL dataset by considering the following criteria: 1) minimum variance of 0.06, 2) minimum correlation of 0.35 between the two fluorescent probes, and 3) minimum heritable variation of 0.5, where we measure heritable variation with the measure presented by Brem et al.[17]. As previously mentioned, we also chose to model 50 loci. This estimation is derived from assessing the number of loci conducive to 4 or more transcripts at the significance level of 1.0 × 10^{-5} or lower. Next, correlated and adjacent loci were aggregated, yielding 50 genomic "epicenters." This number is roughly consistent with the two recent mapping studies of Zhang et al.[26] and Litvin et al.[24], which identified 25 and 44 loci, respectively.
Network Simulation
Simulating eQTL Data
From the simulated network of 2, 200 transcripts and 50 loci, eQTL data are subsequently generated according to the protocol presented by Zhu et al.[18]. Genotypic data were randomly assigned values for either of the parental strains; genomic loci were assumed to be independent.
The coefficient a_{ i } is drawn from a Gaussian distribution with a mean of 0.75 and standard deviation of 0.2. The error term is drawn from a Gaussian distribution with a mean of 0 and standard deviation of 1. All transcripts will have one or more parents (denoted by x_{ i } ), and the parents may be of any composition of loci and transcripts. The sign of a_{ i } is assigned by drawing from a uniform distribution where the probabilities of positive and negative signs are equal to 0.7 and 0.3, respectively. A fraction (10%) of the transcripts with 2+ parents are selected at random to have non-linear interactions, and are modeled with an additional interaction term: $\mathbf{y}={\displaystyle \sum _{i}^{N}{a}_{i}}{\mathbf{x}}_{\mathbf{i}}+{\displaystyle \sum _{i,j=1}^{N}{k}_{i,j}}{\mathbf{x}}_{\mathbf{i}}{\mathbf{x}}_{\mathbf{j}}+{\u03f5}_{\mathbf{i}}$
For these traits, the interaction term, k_{i,j}, is drawn from a Gaussian distribution with a mean of 0.5 and a standard deviation of 0.1. Ultimately, the mean correlation between parent and child is 0.68. To generate eQTL data composed of weaker correlation structure, we drew a_{ i } from a Gaussian distribution with a mean of 0.6 and a standard deviation of 0.2. The network composed of weaker linkages has a mean correlation between parent and child of 0.55. Table 1 summarizes the properties of the eQTL datasets (rows 2-4), and the parameters used to generate the data (rows 5-8).
In summary, we generated both strongly and weakly correlated datasets, each composed of 100, 200, and 300 samples, resulting in six total datasets.
Traditional Univariate eQTL Mapping
We used the t-test statistic to implement traditional univariate eQTL mapping, which involves an exhaustive search between all transcripts and loci. For each transcript-locus test, the expression levels for the transcript across all segregants are partitioned by the genotypes at the locus. Subsequently, the t-test is performed to assess the extent to which a locus influences the expression level of a transcript. This is repeated for all transcripts against all loci. In order to account for multiple hypothesis testing, we applied the false discovery rate (FDR) test of Benjamini et al.[34]. For eQTL mapping, this entails permuting the expression data prior to measuring the t-test statistic [35]. This shuffling procedure is repeated 10, 000 times to simulate the null distribution, and the null statistics are pooled across transcripts in order to obtain the "experimentwise" threshold [35]. Given this information, the FDR for a nominal p-value can be expressed as: $\text{FDR}=\frac{\text{false}\phantom{\rule{0.5em}{0ex}}\phantom{\rule{0.5em}{0ex}}\text{linkages}}{\text{true}\phantom{\rule{0.5em}{0ex}}\phantom{\rule{0.5em}{0ex}}\text{linkages}}$, where false linkages are defined as instances where the null hypothesis is falsely rejected.
0.1 LCMS Method
Each model applies to a locus, L_{ j } , and two transcripts, t_{ x } and t_{ y } . Each likelihood function iterates over the N possible experiments and the 2 possible states of L_{ j } . Likelihood functions of the form $l({\theta}_{{t}_{{x}_{i}}|{L}_{j}};{t}_{{x}_{i}}|{L}_{j})$ are implemented as a bivariate normal distribution, whereas the likelihood function in M3, $l({\theta}_{{t}_{{y}_{i}}|{t}_{{x}_{i}},{L}_{j}};{t}_{{y}_{i}}|{t}_{{x}_{i}},{L}_{j})$, is modeled as a conditional bivariate normal distribution. The complete functions are outlined in the supplemental material section of Schadt et al.[12].
Bootstrapping is applied 1, 000 times for each triplet, from which the probability of each of the respective models is obtained. Given these probabilities, the actual transcript-transcript priors are obtained by the following rules:
To summarize this logic, the authors prefer to downweight the prior score over two transcripts in cases where the independent model has a probability greater than 0.5.
Stochastic Causal Tree Method
There are several functions that could reasonably be used to implement the potentials, including Pearson's correlation, mutual information, or regression functions. We considered both the PCC and regression functions for our study. Ultimately, due to the fact that eQTL datasets consist of both binary (loci) and continuous (expression) data, we opted to employ linear regression functions to model the potentials. In addition to being suitable for modeling datasets composed of binary and continuous variables, regression functions lend adaptability to studies involving diploids where heterozygosity can be represented by a separate category from either homozygous state. Thus, all figures corresponding to our SCT method in this manuscript are derived from an implementation with regression functions. However, for reference we include a performance comparison between the PCC and regression functions in Additional file 1, figure S 1. We found that both approaches yield comparable results. For the second-order potential, ϕ(n_{ p } , n_{ c } ), we use the following function:
1. Let e be the residuals from the linear regression model: n_{c} = β_{0} + β_{1} * n_{p} + ϵ
2. Set $\varphi ({n}_{g},{n}_{c}|{n}_{p})=\sqrt{SS{R}_{{n}_{g},e}/SST{O}_{{n}_{g},e}}=\sqrt{{R}_{{n}_{g},e}^{2}}$, where $SS{R}_{{n}_{g},e}$ and $SST{O}_{{n}_{g},e}$ are the regression sum of squares and total sum of squares, respectively, from the linear regression model: e = β_{0} + β_{1} * n_{ g } + ε If n_{ p } is an intermediate between n_{ g } and n_{ c } , then the residuals from the first step will not result in a high value for $\varphi ({n}_{g},{n}_{c}|{n}_{p})=\sqrt{SS{R}_{{n}_{g},e}/SST{O}_{{n}_{g},e}}=\sqrt{{R}_{{n}_{g},e}^{2}}$ from the second linear regression function. In other words, n_{ g } will not be able to predict anything significant about n_{ c } , given that n_{ c } has already been regressed against n_{ p } . We use the ordinary least squares (OLS) method to estimate the beta parameters [36].
The covariance, ρ, is assumed to be zero in our study. Also, genomic loci are restricted to the explanatory variable when obtaining the sampling distributions. Note that for two continuous variables, the $\sqrt{{R}^{2}}$ value is symmetric and equal to the correlation coefficient.
where c_{ 1 } is a constant used to modulate the extent to which the conditional potential is weighted against the pairwise potential. We discuss the optimization of this constant in the "Parameter Optimization" subsection. Finally, we note that equation 5 represents a likelihood model and is not a formal decomposition of the joint probability, in which case the probability of the causal chain would be represented by (n_{ g } ) * (n_{ p }|n_{ g } ) (n_{ c } |n_{ p } ).
The SCT algorithm will stochastically choose between the five optimal transitions. The choice is weighted by a factor of the log likelihood score to a power represented by parameter p. Higher values of p bias the stochastic choice towards higher log likelihood scores. We discuss the importance of this weighting in the "Parameter Optimization" subsection. This process is repeated until the maximum number of nodes allowed per tree is reached, represented by the parameter M. We set M = 60, making it likely that the trees cover a number of transcripts equal to the ratio of transcripts to loci. We discuss how performance is affected by varying values of M in the "Parameter Optimization" subsection. Finally, we obtain ϕ(n_{ g } , n_{ c }|n_{ p } ) for the root nodes by randomly generating a value according to a Gaussian distribution with parameters μ equal to 0.0 and σ equal to the value obtained via the sampling procedure described in Additional file 1 table S1. In practice, we found that simulating random values for the conditional potential is as effective as optimizing a fixed parameter.
The SCT method produces a set of trees which can be represented as adjacency matrices. The SCT output is converted into an n × n prior matrix by dividing the frequency at which node x is a parent of node y by the number of times the SCT method is run. For example, we ran the method 1, 000 times from each of the 50 loci. If node x is a parent of node y on 100 occasions, then the prior matrix will have a value of $\frac{100}{50,000}$ entry x, y, since there are 50 * 1, 000 trees constructed.
Formally, the SCT method is described as follows:
The dataset D is composed of two components, gene expression and genotypic data. We represent the expression data E as the set of N variables (genes) x_{ i } , ... x_{ N } ∈ X. Similarly, the genotypic dataset G consists of the set of M variables (loci) L_{ i } ,...L_{ M } ∈ G. There are P samples (segregants), and therefore the two components of the dataset D can be expressed as: E = x[h]....x[P] and G = g[h]....g[P]. These data represent the input to the algorithm.
Input : The dataset, D = {d[1], ..., d[P]}
Input : M, the maximum number of leaves on a tree
Input : I, the number of iterations for the procedure (conducted over all loci)
Input : L, the set of l loci
Output: R, the N × N prior matrix representing the causal relationships between nodes
for i ← 1 to I do
for l ∈ L do
S←∅;// S is the set nodes included in the tree
include node l into the set S;
for m ← 1 to M do
leaf ← StochasticLeaf(S);
S ← S∪ {leaf}; // add leaf to the set, S
m ← m + 1;
i ←i + 1
// Convert the trees into the prior matrix, R.
Algorithm 1: SCT Main Procedure
Input : S, the set of nodes currently included in the tree
Output: the next leaf to be included in the tree
// B is the set of best candidates
B←∅;
for s ∈ S do
b = GetBestTransition(s.id, s. parentId);
B ← B∪{b};
return b ∈ B, where b is stochastically chosen based on relative scores.
Algorithm 2: StochasticLeaf (Node[] S)
Input : p, the leaf node
Input : g, the parent of p
Output: the optimal transition corresponding to p and g;
// C is the set of candidates for node p
for c ∈ C do
score_{g,p,c}= c_{1}log(l(ϕ(g, c|p))) - log(l(ϕ(p, c)))
return c_{ best } ∈ C, where c_{ best } corresponds to max(score_{g,p,*});
Algorithm 3: GetBestTransition (int p, int g)
Bayesian Network Structure Learning
Bayesian networks provide a graphical representation of the joint probability distribution for a set of random variables, allowing for efficient computation of the probability of graphical structures [19]. A signature aspect of Bayesian networks is that the global likelihood function that defines the probability of the structure is decomposable into the product of local likelihood functions. The local likelihood score for each node can be computed based on the sufficient statistics encoded by its parents. Formally, Bayes' formula allows the posterior probability of a network, P(G|D), to be calculated as P(G) * P(D|G). Since the number of possible network structures grows super-exponentially with the number of nodes, it is not possible to conduct an exhaustive search over the space of possible networks. In light of this practical limitation, we use Markov Chain Monte Carlo (MCMC) simulations to explore the space of high-scoring network structures [38] from which we sample to construct a consensus network.
To improve the computational feasibility of the structure learning algorithm, we restrict the maximum number of parents to be three, a constraint that is employed by several other studies [13, 18, 32, 38, 39]. We further confine the search space by only considering edges between a node and a predefined set of candidates for each node. To construct the candidate set, we initially calculate each node's top-k associations. We next traverse the top-k sets to ensure reciprocity. For example, if node y is a top-k candidate of node x, we establish node x as a candidate of node y, if it wasn't already. This symmetry ensures that the transition matrix is reversible. For this study, we chose k = 10, which results in roughly 15 candidates per node once reciprocity is considered (i.e. the candidate set is expanded). This relatively stringent candidate set is chosen for computational efficiency, but it does encompass nearly 97% of the true edges for each of the six datasets.
New structures are drawn from the proposal distribution, Q(G; G'), and the proposed structure G' is randomly chosen from the set of legal moves corresponding to G, which is denoted as η(G). η(G) is constrained by the following criteria: 1) the move involves the addition, removal, or reversal of an existing edge; 2) the move does not result in a cycle; 3) the move does not result in any node having more than three parents; and 4) the move involves a candidate node. One subtle aspect of the proposal distribution is that is not necessarily symmetric [32, 38]. Therefore, η(G) and η(G') must be explicitly calculated and accounted for in the acceptance function. The computation of the Hastings factor, Q(G; G')/Q(G'; G), thus reduces to the ratio of cardinalities of the two sets: $Q(G;{G}^{\prime})/Q({G}^{\prime};G)=\frac{|\eta (G)|}{|\eta ({G}^{\prime})|}$. We note that maintaining η(G) is made computationally feasible by dynamically tracking node ancestries during the simulations.
Where e(G) denotes the set of edges in G.
For each of the three structure learning methods, and for each dataset, we initiate two parallel MCMC runs. Each run consists of 150 million iterations with a burn-in period of 10 million iterations. We assess convergence in the subsequent 140 million iterations. As has been noted by others [38, 39, 41], a necessary but not sufficient property for convergence is the apparent stabilization of edge frequencies. Thus a simple heuristic to assess convergence involves measuring the distortion between sampling intervals, defined as S_{ i } , where i indicates the i^{ th } interval. We sample a network at every 200 iterations, leading to 140M/200 = 700, 000 total networks sampled in each run. Upon convergence, the distortion in edge frequencies between sampling intervals should converge, both within and between MCMC simulations. Let S_{ A } be the set of sampling intervals for the first MCMC simulation, and let S_{ B } be the set for the second MCMC simulation. Upon convergence, the intra-run distortions, D(S_{A,i}, S_{A,i+1}) and D(S_{ B,i }, S_{B,i+1}), should converge to the same level as the inter-run distortions, D(S_{ A,i } , S_{B,i}). We define distortion as the difference in edge frequencies over a sampling interval. For example, if an interval consists of 100 samples, and if edge i, j occurs in $\frac{90}{100}$ samples for interval A and in $\frac{70}{100}$ samples during interval B, then the distortion for edge i, j is 0.9 - 0.7 = 0.2. To obtain the distortion for a graph, we average over the distortions of all possible edges. We found that 150M iterations was sufficient to demonstrate that the distortion between the two MCMC runs is reduced to 0.005. Additional file 1 figure S 2 illustrates convergence for all three methods when applied to the dataset with weaker correlation structure and 200 samples. Specifically, figure S 2 shows the relationship between performance (AUC) and convergence (distortion) as the number of iterations increases. It is clear that performance (AUC) reaches a plateau after 20M iterations (log(20M) = 7.3), indicating that, although edge distortion will in theory reach 0.0 over an infinite number of iterations, optimal performance is attained within 150M iterations. Another useful way to illustrate convergence is to plot the edge frequencies from two independent runs. Additional file 1, figures S 3a-S 3c, corresponding to each of the respective methods, indicate strong concurrences in edge frequencies between runs for all three methods.
Beta Parameter; Stronger Correlation
Method | Beta | 100 Samples | 200 Samples | 300 Samples |
---|---|---|---|---|
LCMS | 4 | 0.628 | 0.867 | 0.926 |
LCMS | 8 | 0.646 | 0.856 | 0.925 |
LCMS | 12 | 0.645 | 0.854 | 0.920 |
LCMS | 16 | 0.628 | 0.845 | 0.906 |
LCMS | 20 | 0.616 | 0.831 | 0.894 |
LCMS | 24 | 0.601 | 0.817 | 0.881 |
SCT | 4 | 0.783 | 0.926 | 0.953 |
SCT | 8 | 0.821 | 0.943 | 0.964 |
SCT | 12 | 0.837 | 0.953 | 0.969 |
SCT | 16 | 0.843 | 0.959 | 0.967 |
SCT | 20 | 0.841 | 0.958 | 0.968 |
SCT | 24 | 0.842 | 0.957 | 0.969 |
Beta Parameter; Weaker Correlation
Method | Beta | 100 Samples | 200 Samples | 300 Samples |
---|---|---|---|---|
LCMS | 4 | 0.577 | 0.787 | 0.884 |
LCMS | 8 | 0.578 | 0.788 | 0.884 |
LCMS | 12 | 0.587 | 0.781 | 0.886 |
LCMS | 16 | 0.574 | 0.786 | 0.885 |
LCMS | 20 | 0.574 | 0.773 | 0.880 |
LCMS | 24 | 0.562 | 0.773 | 0.881 |
SCT | 4 | 0.689 | 0.847 | 0.925 |
SCT | 8 | 0.715 | 0.869 | 0.935 |
SCT | 12 | 0.726 | 0.878 | 0.939 |
SCT | 16 | 0.726 | 0.883 | 0.940 |
SCT | 20 | 0.724 | 0.885 | 0.941 |
SCT | 24 | 0.725 | 0.884 | 0.939 |
The continuous data are discretized via k-means clustering into three levels representing down-regulated expression, steady expression, and up-regulated expression. The data are parameterized with a multinomial distribution and we utilized the structure-equivalent Dirichlet priors introduced by Heckerman et al.[43].
Therefore, while both the LCMS and SCT methods utilize continuous data to produce a prior matrix, the Bayesian network structure learning procedure uses discrete data. We preferred the computational efficiency associated with discrete data given the large number of simulations conducted in our study. Finally, we note that genomic loci are modeled as head variables in our networks, which has two consequences: 1) we do not need to test for cis- versus trans-relationships; 2) We are able to use the resulting networks for expression quantitative trait loci (eQTL) mapping.
Parameter Optimization
c1 and p Parameter Optimization; Stronger Correlation; 200 Samples
p | |||||
---|---|---|---|---|---|
1.0 | 2.0 | 3.0 | 4.0 | ||
c _{ 1 } | 0.0 | 0.606 | 0.668 | 0.672 | 0.660 |
1.0 | 0.661 | 0.740 | 0.751 | 0.751 | |
2.0 | 0.686 | 0.777 | 0.801 | 0.790 | |
3.0 | 0.718 | 0.804 | 0.824 | 0.812 | |
4.0 | 0.736 | 0.823 | 0.839 | 0.822 | |
5.0 | 0.753 | 0.836 | 0.850 | 0.857 | |
6.0 | 0.771 | 0.841 | 0.854 | 0.836 |
c1 and p Parameter Optimization; Weaker Correlation; 200 Samples
p | |||||
---|---|---|---|---|---|
1.0 | 2.0 | 3.0 | 4.0 | ||
c _{ 1 } | 0.0 | 0.600 | 0.652 | 0.643 | 0.635 |
1.0 | 0.630 | 0.692 | 0.698 | 0.668 | |
2.0 | 0.662 | 0.722 | 0.726 | 0.700 | |
3.0 | 0.677 | 0.739 | 0.746 | 0.722 | |
4.0 | 0.698 | 0.772 | 0.762 | 0.738 | |
5.0 | 0.714 | 0.768 | 0.778 | 0.750 | |
6.0 | 0.727 | 0.798 | 0.786 | 0.782 |
Precision-recall curves
To assess the quality of the reconstructed networks, we generated precision-recall plots, which provide a graphical depiction of precision versus recall. Precision is defined as $\text{precision}=\frac{\ne \text{true}\phantom{\rule{0.5em}{0ex}}\text{positives}}{\ne \text{true}\phantom{\rule{0.5em}{0ex}}\text{positives}+\ne \text{false}\phantom{\rule{0.5em}{0ex}}\text{positives}}$, and recall is defined as $\text{recall}=\frac{\ne \text{true}\phantom{\rule{0.5em}{0ex}}\text{positives}}{\ne \text{total}\phantom{\rule{0.5em}{0ex}}\text{true}\phantom{\rule{0.5em}{0ex}}\text{positives}}$. Precision indicates the sensitivity of the method, whereas recall indicates the extent to which the method can detect true edges. Higher levels of precision and recall are both desirable, and thus so are higher values of area under the curve (AUC).
For causal edge detection, we construct precision-recall curves by sampling 1:4 M networks at every 200 iterations over an interval consisting of 140 M iterations for two individual MCMC runs. Therefore, the total number of networks sampled is expressed as: $2\phantom{\rule{0.5em}{0ex}}\text{runs}*\frac{140\phantom{\rule{0.5em}{0ex}}\text{M}\phantom{\rule{0.5em}{0ex}}\text{iterations}}{\text{run}}*\frac{1\phantom{\rule{0.5em}{0ex}}\text{network}}{200\phantom{\rule{0.5em}{0ex}}\text{iterations}}=1.4\text{M}$ networks.
Though far fewer networks would likely be sufficient to assess performance, we note that at least 10, 000 thousand samples are ideal, given that any sampled network from the MCMC chain will contain many suboptimal edges. To generate precision-recall curves for the consensus network, we lower the threshold frequency at or above which an edge occurs, starting at 1.4 M. If an edge occurs at or above the threshold, it is assigned a true or false positive based on whether the edge is present in the true network. The threshold is repeatedly decremented by 1 until reaching 0.
For eQTL mapping, we take individual networks and establish transcript-loci (eQTL) linkages by conducting a depth-first from each locus. That is, an eQTL linkage is established if there is a directed path from a locus to a transcript. Since any sampled network from the MCMC chain will contain a considerable number of extraneous edges, we extract 1, 000 networks that are spaced 5, 000 iterations apart, then use the extracted networks as initializations for greedy optimization procedures [44] with 200, 000 iterations. This results in the removal of suboptimal edges that are deleterious to eQTL mapping. In summary, for each method and for each of the six datasets, we greedily optimized 1, 000 networks upon which we conducted depth-first searches to establish eQTL linkages. The depth-first searches on the individual networks yields 1, 000 sets of transcript-locus linkages. We generate precision-recall curves by lowering the threshold frequency at or above which a transcript-locus linkage occurs, starting at 1, 000 and decrementing by 1 until reaching 0. True eQTL linkages are established by running depth-first searches on the true network.
Availability
Software for the SCT method and MCMC Bayesian structure learning procedure can be accessed online: http://www.cs.ucsb.edu/%7Edbl/software.php
Declarations
Acknowledgements
This work is supported in part by National Science Foundation grants IIS-0917149 and IIS-0612327. We thank Jarad Niemi for providing insight on the subject of stepwise regression. We also thank two anonymous reviewers for advice in several areas, including QTL mapping and MCMC methods.
Authors’ Affiliations
References
- Schena M, Shalon D, Davis RW, Brown PO: Quantitative monitoring of gene expression patterns with a complementary DNA microarray. Science 1995, 270: 467–470. 10.1126/science.270.5235.467View ArticlePubMedGoogle Scholar
- Lockhart DJ, Dong H, Byrne MC, Follettie MT, Gallo MV, Chee MS, Mittmann M, Wang C, Kobayashi M, Horton H, Brown EL: Expression monitoring by hybridization to high-density oligonucleotide arrays. Nat Biotechnol 1996, 14: 1675–1680. 10.1038/nbt1296-1675View ArticlePubMedGoogle Scholar
- Wang Z, Gerstein M, Snyder M: RNA-Seq: a revolutionary tool for transcriptomics. Nat Rev Genet 2009, 10: 57–63. 10.1038/nrg2484PubMed CentralView ArticlePubMedGoogle Scholar
- Jansen RC, Nap JP: Genetical genomics: the added value from segregation. Trends Genet 2001, 17: 388–391. 10.1016/S0168-9525(01)02310-1View ArticlePubMedGoogle Scholar
- Brem RB, Yvert G, Clinton R, Kruglyak L: Genetic Dissection of Transcriptional Regulation in Budding Yeast. Science 2002, 296(5594):752–755. 10.1126/science.1069516View ArticlePubMedGoogle Scholar
- Schadt EE, Monks SA, Drake TA, Lusis AJ, Che N, Colinayo V, Ruff TG, Milligan SB, Lamb JR, Cavet G, Linsley PS, Mao M, Stoughton RB, Friend SH: Genetics of gene expression surveyed in maize, mouse and man. Nature 2003, 422: 297–302. 10.1038/nature01434View ArticlePubMedGoogle Scholar
- Morley M, Molony CM, Weber TM, Devlin JL, Ewens KG, Spielman RS, Cheung VG: Genetic analysis of genome-wide variation in human gene expression. Nature 2004, 430: 743–747. 10.1038/nature02797PubMed CentralView ArticlePubMedGoogle Scholar
- Stranger BE, Forrest MS, Clark AG, Minichiello MJ, Deutsch S, Lyle R, Hunt S, Kahl B, Antonarakis SE, Tavare S, Deloukas P, Dermitzakis ET: Genome-wide associations of gene expression variation in humans. PLoS Genet 2005, 1: e78. 10.1371/journal.pgen.0010078PubMed CentralView ArticlePubMedGoogle Scholar
- Keurentjes JJ, Fu J, Terpstra IR, Garcia JM, van den Ackerveken G, Snoek BL, Peeters AJ, Vreugdenhil D, Koornneef M, Jansen RC: Regulatory network construction in Arabidopsis by using genome-wide gene expression quantitative trait loci. PNAS 2007, 104(5):1708–1713. 10.1073/pnas.0610429104PubMed CentralView ArticlePubMedGoogle Scholar
- Dimas AS, Stranger BE, Beazley C, Finn RD, Ingle CE, Forrest MS, Ritchie ME, Deloukas P, Tavare S, Dermitzakis ET: Modifier effects between regulatory and protein-coding variation. PLoS Genet 2008, 4: e1000244. 10.1371/journal.pgen.1000244PubMed CentralView ArticlePubMedGoogle Scholar
- Zhu J, Lum PY, Lamb J, GuhaThakurta D, Edwards SW, Thieringer R, Berger JP, Wu MS, Thompson J, Sachs AB, Schadt EE: An integrative genomics approach to the reconstruction of gene networks in segregating populations. Cytogenet Genome Res 2004, 105(2–4):363–374. 10.1159/000078209View ArticlePubMedGoogle Scholar
- Schadt EE, Lamb J, Yang X, Zhu J, Edwards S, Guhathakurta D, Sieberts SK, Monks S, Reitman M, Zhang C, Lum PY, Leonardson A, Thieringer R, Metzger JM, Yang L, Castle J, Zhu H, Kash SF, Drake TA, Sachs A, Lusis AJ: An integrative genomics approach to infer causal associations between gene expression and disease. Nature Genetics 2005, (7):710–717. 10.1038/ng1589Google Scholar
- Zhu J, Zhang B, Smith ENN, Drees B, Brem RBB, Kruglyak L, Bumgarner REE, Schadt EEE: Integrating large-scale functional genomic data to dissect the complexity of yeast regulatory networks. Nature genetics 2008, 40: 854–861. 10.1038/ng.167PubMed CentralView ArticlePubMedGoogle Scholar
- Bing N, Hoeschele I: Genetical genomics analysis of a yeast segregant population for transcription network inference. Genetics 2005, 170(2):533–542. 10.1534/genetics.105.041103PubMed CentralView ArticlePubMedGoogle Scholar
- Chaibub Neto E, Ferrara CT, Attie AD, Yandell BS: Inferring causal phenotype networks from segregating populations. Genetics 2008, 179: 1089–1100. 10.1534/genetics.107.085167PubMed CentralView ArticlePubMedGoogle Scholar
- Kulp DC, Jagalur M: Causal inference of regulator-target pairs by gene mapping of expression profiles. BMC Genomics 2006., 7: 10.1186/1471-2164-7-125Google Scholar
- Brem RB, Kruglyak L: The landscape of genetic complexity across 5,700 gene expression traits in yeast. PNAS 2005, 102(5):1572–1577. 10.1073/pnas.0408709102PubMed CentralView ArticlePubMedGoogle Scholar
- Zhu J, Wiener MC, Zhang C, Fridman A, Minch E, Lum PY, Sachs JR, Schadt EE: Increasing the power to detect causal associations by combining genotypic and expression data in segregating populations. PLoS Comput Biol 2007, 3: e69. 10.1371/journal.pcbi.0030069PubMed CentralView ArticlePubMedGoogle Scholar
- Pearl J: Probabilistic reasoning in intelligent systems: networks of plausible inference. San Francisco, CA, USA: Morgan Kaufmann Publishers Inc; 1988.Google Scholar
- Doerge RW: Mapping and analysis of quantitative trait loci in experimental populations. Nat Rev Genet 2002, 3: 43–52. 10.1038/nrg703View ArticlePubMedGoogle Scholar
- Storey JD, Akey JM, Kruglyak L: Multiple Locus Linkage Analysis of Genomewide Expression in Yeast. PLoS Biology 2007, 3: e267+. 10.1371/journal.pbio.0030267View ArticleGoogle Scholar
- Zou W, Zeng ZB: Multiple interval mapping for gene expression QTL analysis. Genetica 2009, 137: 125–134. 10.1007/s10709-009-9365-zView ArticlePubMedGoogle Scholar
- Lee SI, Dudley AM, Drubin D, Silver PA, Krogan NJ, Pe'er D, Koller D: Learning a prior on regulatory potential from eQTL data. PLoS Genet 2009, 5: e1000358. 10.1371/journal.pgen.1000358PubMed CentralView ArticlePubMedGoogle Scholar
- Litvin O, Causton HC, Chen BJ, Pe'er D: Modularity and interactions in the genetics of gene expression. Proc Natl Acad Sci USA 2009, 106: 6441–6446. 10.1073/pnas.0810208106PubMed CentralView ArticlePubMedGoogle Scholar
- Pan W: Network-based multiple locus linkage analysis of expression traits. Bioinformatics 2009, 25: 1390–1396. 10.1093/bioinformatics/btp177PubMed CentralView ArticlePubMedGoogle Scholar
- Zhang W, Zhu J, Schadt EE, Liu JS: A Bayesian partition method for detecting pleiotropic and epistatic eQTL modules. PLoS Comput Biol 2010, 6: e1000642. 10.1371/journal.pcbi.1000642PubMed CentralView ArticlePubMedGoogle Scholar
- Perlstein EO, Ruderfer DM, Roberts DC, Schreiber SL, Kruglyak L: Genetic basis of individual differences in the response to small-molecule drugs in yeast. Nat Genet 2007, 39: 496–502. 10.1038/ng1991View ArticlePubMedGoogle Scholar
- Chen BJ, Causton HC, Mancenido D, Goddard NL, Perlstein EO, Pe'er D: Harnessing gene expression to identify the genetic basis of drug resistance. Mol Syst Biol 2009, 5: 310.PubMed CentralPubMedGoogle Scholar
- Ruderfer DM, Roberts DC, Schreiber SL, Perlstein EO, Kruglyak L: Using expression and genotype to predict drug response in yeast. PLoS ONE 2009, 4: e6907. 10.1371/journal.pone.0006907PubMed CentralView ArticlePubMedGoogle Scholar
- Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM, K Dolinski, Dwight SS, Eppig JT, Harris MA, Hill DP, Issel-Tarver L, Kasarskis A, Lewis S, Matese JC, Richardson JE, Ringwalk M, Rubin GM, Sherlock G: Gene Ontology: tool for the unification of biology. The gene ontology consortium. Nature Biotechnology 2000, 25: 25–29.Google Scholar
- Imoto S, Higuchi T, Goto T, Tashiro K, Kuhara S, Miyano S: Combining microarrays and biological knowledge for estimating gene networks via Bayesian networks. Proc IEEE Comput Soc Bioinform Conf 2003, 2: 104–113.PubMedGoogle Scholar
- Husmeier D, Werhli AV: Bayesian integration of biological prior knowledge into the reconstruction of gene regulatory networks with Bayesian networks. Comput Syst Bioinformatics Conf 2007, 6: 85–95.View ArticlePubMedGoogle Scholar
- Werhli AV, Husmeier D: Gene regulatory network reconstruction by Bayesian integration of prior knowledge and/or different experimental conditions. J Bioinform Comput Biol 2008, 6: 543–572. 10.1142/S0219720008003539View ArticlePubMedGoogle Scholar
- Benjamini Y, Hochberg Y: Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society. Series B (Methodological) 1995, 57: 289–300.Google Scholar
- Churchill GA, Doerge RW: Empirical threshold values for quantitative trait mapping. Genetics 1994, 138: 963–971.PubMed CentralPubMedGoogle Scholar
- Kutner MH, Nachtsheim CJ, Neter J: Applied Linear Regression Models. fourth international edition. McGraw-Hill/Irwin; 2004.Google Scholar
- Faith JJ, Hayete B, Thaden JT, Mogno I, Wierzbowski J, Cottarel G, Kasif S, Collins JJ, Gardner TS: Large-scale mapping and validation of Escherichia coli transcriptional regulation from a compendium of expression profiles. PLoS Biol 2007, 5: e8. 10.1371/journal.pbio.0050008PubMed CentralView ArticlePubMedGoogle Scholar
- Madigan D, York J: Bayesian graphical models for discrete data. Intl Stat Rev 1995, 63: 215–232. 10.2307/1403615View ArticleGoogle Scholar
- Friedman N, Koller D: Being Bayesian about Network Structure. In UAI '00: Proceedings of the 16th Conference on Uncertainty in Artificial Intelligence. San Francisco, CA, USA: Morgan Kaufmann Publishers Inc; 2000:201–210.Google Scholar
- Mukherjee S, Speed TP: Network inference using informative priors. Proc Natl Acad Sci USA 2008, 105: 14313–14318. 10.1073/pnas.0802272105PubMed CentralView ArticlePubMedGoogle Scholar
- Husmeier D, Dybowski R, Roberts S: Probabilistic Modelling in Bioinformatics and Medical Informatics. Secaucus, NJ, USA: Springer-Verlag New York, Inc; 2004.Google Scholar
- Smith AFM, Roberts GO: Bayesian Computation Via the Gibbs Sampler and Related Markov Chain Monte Carlo Methods. Journal of the Royal Statistical Society (Series B) 1993., 55:Google Scholar
- Heckerman D, Geiger D: Learning Bayesian networks: A unification for discrete and Gaussian domains. Proceedings of the Eleventh Annual Conference on Uncertainty in Artificial Intelligence 1995, 274–284.Google Scholar
- Friedman N, Linial M, Nachman I: Using Bayesian networks to analyze expression data. Journal of Computational Biology 2000, 7: 601–620. 10.1089/106652700750050961View ArticlePubMedGoogle Scholar
Copyright
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.