- Research article
- Open Access
An intuitionistic approach to scoring DNA sequences against transcription factor binding site motifs
- Fernando Garcia-Alcalde^{1, 2}Email author,
- Armando Blanco^{2} and
- Adrian J Shepherd^{3}
https://doi.org/10.1186/1471-2105-11-551
© Garcia-Alcalde et al; licensee BioMed Central Ltd. 2010
Received: 27 April 2010
Accepted: 8 November 2010
Published: 8 November 2010
Abstract
Background
Transcription factors (TFs) control transcription by binding to specific regions of DNA called transcription factor binding sites (TFBSs). The identification of TFBSs is a crucial problem in computational biology and includes the subtask of predicting the location of known TFBS motifs in a given DNA sequence. It has previously been shown that, when scoring matches to known TFBS motifs, interdependencies between positions within a motif should be taken into account. However, this remains a challenging task owing to the fact that sequences similar to those of known TFBSs can occur by chance with a relatively high frequency. Here we present a new method for matching sequences to TFBS motifs based on intuitionistic fuzzy sets (IFS) theory, an approach that has been shown to be particularly appropriate for tackling problems that embody a high degree of uncertainty.
Results
We propose SC_{ intuit }, a new scoring method for measuring sequence-motif affinity based on IFS theory. Unlike existing methods that consider dependencies between positions, SC_{ intuit } is designed to prevent overestimation of less conserved positions of TFBSs. For a given pair of bases, SC_{ intuit } is computed not only as a function of their combined probability of occurrence, but also taking into account the individual importance of each single base at its corresponding position. We used SC_{ intuit } to identify known TFBSs in DNA sequences. Our method provides excellent results when dealing with both synthetic and real data, outperforming the sensitivity and the specificity of two existing methods in all the experiments we performed.
Conclusions
The results show that SC_{ intuit } improves the prediction quality for TFs of the existing approaches without compromising sensitivity. In addition, we show how SC_{ intuit } can be successfully applied to real research problems. In this study the reliability of the IFS theory for motif discovery tasks is proven.
Keywords
- Membership Degree
- Invasive Pulmonary Aspergillosis
- Positional Dependency
- Position Weighted Matrix
- Transcription Factor Binding Site Motif
Background
Cells control the abundance of proteins by means of diverse mechanisms. One such mechanism is the regulation of transcription, which is a continuous process whereby many factors combine to ensure appropriate rates of protein synthesis. Understanding such complex processes is one of the main objectives in computational biology. In its early stages, transcription is controlled, among other mechanisms, by the binding of proteins called transcription factors (TFs) to specific regions of a given chromosome called transcription factor binding sites (TFBSs). These interactions between proteins and DNA usually take place upstream from the gene, close to the transcription start site (TSS), in the so-called promoter region of the gene.
One of the biggest issues in identifying TFBSs is that a single binding protein can bind to different DNA sequences. Related DNA sequences to which the same TF can bind are grouped together into a TFBS motif. The identification of TFBSs within a given set of DNA sequences is an active area of research. In this context there exist two main approaches: i) the de novo discovery of motifs, and ii) the detection of TFBSs using motifs that are already known.
De novo methods aim to find significant sub-sequence patterns within a set of TFBS sequences. Some of the most popular approaches are MEME [1], Gibbs sampling [2], AlignACE [3], and more recently PRIORITY and Trawler [4, 5]. For a review see [6].
Detection methods, on the other hand, focus on inferring new TFBSs from known binding motifs. Early detection methods assumed independence between positions within a putative TFBS sequence, e.g. in Patser [7] and ConSite [8]. However, it is now well established that this assumption is wrong [9–11], and some methods that consider position dependency for modeling and finding TFBSs using advanced HMM and Bayesian models have appeared [12, 13]. Likewise, two recent detection methods have been developed that take into account interdependencies between TFBS positions. Tomovic and Oakeley proposed a method that incorporates a measure of positional interdependence into the overall score [14]. More recently, Zare-Mirakabad et al. developed a method based on joint information content and mutual information [15]. In this method, positional dependencies are taken into account by considering all pairwise combinations of positions (see the Methods section for more information).
The fact that TFBS sequences are usually very short means that the same or very similar sequences tend to occur by chance at a relatively high frequency. Consequently one of the main goals in the prediction of TFBSs is to reduce the false positive rate without compromising sensitivity. Methods that take into account positional dependencies tend to be significantly more effective at meeting this challenge. However, there remains room for improvement. As we will show in the Results section, existing methods have some drawbacks, such as overlearning of the training data, arbitrary threshold selection for testing dependencies, etc. The purpose of the work presented here is to provide a new method for measuring sequence-motif affinity that improves on existing approaches.
Zadeh proposed fuzzy sets theory to mathematically model the imprecision inherent in certains concepts [16]. Briefly, fuzzy sets theory allows an object to partially belong to a set with a membership degree between 0 and 1. Classical set theory is a special case of its fuzzy counterpart in which membership and certainty degrees are restricted to either 0 or 1. Atanassov proposed intuitionistic fuzzy sets (IFS) theory as an extension of the fuzzy sets theory [17]. IFSs generalize the notion of a fuzzy set representing uncertainty with respect to both the degree of membership (μ) and non-membership (ν) of a set by allowing that the sum μ + ν ≤ 1.
Owing to the fact that IFSs are capable of modelling the uncertainty present in real-life situations, they have been widely applied during the past decades to a variety of problems (see the Methods section). In recent years, it has been seen that the inherent uncertainty and noise that characterize biological data cannot always be modeled sufficiently well using probabilistic approaches and that, as a consequence, alternative approaches to modelling this uncertainty may be required [18–21]. In addition to the usual problems of missing values and noisy data associated with biological data, there exist some additional hidden factors that affect binding affinities in the context of sequence-motif scoring, e.g. cooperative binding and chromatin structure [22]. Furthermore, the described motifs are subject to change as new experiments confirm new binding sites. In this work we make use of IFS theory to formally model the uncertainty associated with the problem of scoring DNA sequences against TFBS motifs.
Results
Case studies
First, we wanted to show the ability of our proposed method, SC_{ intuit }, to discriminate between the relative importance of poorly-conserved positions and well-conserved positions comparing it with the most representative scoring methods: i) SC_{ indep }, a probabilistic method that assumes positional independence; ii) SC_{ dep }, a scoring method proposed by Tomovic and Oakeley that take into account statistical interdependencies between TFBS positions [14]; and iii) SC_{ mat }, a scoring function proposed by Zare-Mirakabad et al. based on the dependency between all pairwise combinations of binding site positions [15].
Scoring results
Non-conserved | Conserved | Difference | |
---|---|---|---|
SC _{ intuit } | 0.788 | 0.687 | 0.101 |
SC _{ dep } | 0.832 | 0.815 | 0.017 |
SC _{ mat } | 0.672 | 0.685 | -0.013 |
SC _{ indep } | 0.839 | 0.827 | 0.012 |
In reality, it would be desirable that the scoring for the case of the mutated outlier sequence be lower than the scoring for the mutated non-oulier sequence. Results obtained by the SC_{ mat }, SC_{ dep }, and SC_{ indep } methods failed to capture the expected differences, giving the incorrect impression that binding is likely to occur. On the other hand, our proposed method obtained a more realistic distance between the sequences, providing a much lower score for the mutated outlier sequence (Figure 2(C)).
These insights are confirmed in the following sections where the experiments are extended to use large datasets, and the results are measured in terms of discovery rates.
Prediction of TFBSs
Synthetic sequences
In order to compare the performance of the different methods in predicting TFBSs, we used the non-redundant publicly available JASPAR motifs database for our experiments [23]. We selected all motifs for which binding sequences are available (not only matrix profiles), resulting in a dataset of 124 motifs. For each of these motifs, a random number between 2 and 6 binding sites were randomly selected and inserted in random sequences of a random length between 200 bp and 500 bp from a third-order Markov model background distribution obtained from the RSAT (Regulatory Sequence Analysis Tools) [24]. For each position of each sequence we computed the score for their corresponding motifs with an assumed known TFBS length (the length of the inserted motif).
Synthetic sequences precision.
Threshold | SC _{ intuit } | SC _{ dep } | SC _{ mat } | SC _{ indep } |
---|---|---|---|---|
0.7 | 0.63 | 0.17 | 0.09 | 0.02 |
0.8 | 0.82 | 0.27 | 0.14 | 0.05 |
AUC values for the synthetic and mutated sequence experiments.
Synthetic | Mutated | |
---|---|---|
SC _{ indep } | 0.550 | 0.526 |
SC _{ dep } | 0.730 | 0.705 |
SC _{ mat } | 0.787 | 0.725 |
SC _{ intuit } | 0.910 | 0.886 |
Mutated sequences
Real Data
True positive rate for the real data experiment
TF | SC _{ intuit } | SC _{ dep } | SC _{ mat } | SC _{ indep } |
---|---|---|---|---|
SMAD1 | 0.96 | 0.94 | 0.90 | 0.86 |
c-Myc | 0.92 | 0.94 | 0.92 | 0.84 |
STAT3 | 0.98 | 0.92 | 0.96 | 0.88 |
Study of Single Nucleotide Polymorphisms in TNFR1 Gene for the Response against Aspergillus Fumigatus
Hematological patients are typically treated by chemotherapy and/or radiation. These treatments usually produce immunosuppression and severe neutropenia. This clinical situation can be exploited by opportunistic pathogens such as Aspergillus fumigatus to cause a deadly infection called Invasive Pulmonary Aspergillosis (IPA) [30, 31]. The importance of finding ways to combat this pathogen is evidenced by the fact that IPA occurs in roughly 10% to 40% of hematological patients, with overall mortality rates ranging from 50% to 90% [32, 33].
Tumor necrosis factor (TNF) activates T lymphocytes in response to fungal infections through TNF receptors. One of the most important TNF receptors is TNFR1, which plays a crucial role in immune regulation and host immune responses. Experimental studies with TNFR1 knockout mice indicate that TNFR1 is indispensable in host resistance against several infections [34]. Our hypothesis is that single nucleotide polymorphisms (SNPs) in the TNFR1 gene may influence the innate immune response against Aspergillus fumigatus.
The gene encoding TNFR1 contain numerous polymorphisms [35, 36]. By means of different experiments, we concluded that TNFR1_{-609(G/T)} polymorphism is critical in the development of the response against Aspergillus because it might be regulating the cell-mediated Th1 immune response. Details on these experiments are out of the scope of this work and can be consulted in [37]. In this section, we use our proposed scoring method SC_{ intuit } to investigate whether the TNFR1_{-609(G/T)} promoter polymorphism is involved in the disruption of the recognition of a potential binding site for a critical transcription factor that could influence TNFR1 transcription level.
TNFR1_{-609(G/T)} Polymorphism Binding Affinity
For this experiment we used TFBS motifs found in TRANSFAC database [29], which has been widely used in research involving regulatory elements [38]. In order to find interesting dependencies between the TNFR1_{-609(G/T)} SNP and TFs binding affinity we scored the human TRANSFAC TFBSs against the TNFR1_{-609(G/T)} polymorphism by means of the SC_{ intuit } method.
SC_{ intuit }scores for the two alleles.
TF | Starting position | Direction | TNFR1-_{609(T)} | TNFR1-_{609(G)} |
---|---|---|---|---|
AREB6 | 603 | - | 0.59 | 0,70 |
E2A | 606 | - | 0.64 | 0.79 |
HNF4 | 605 | + | 0.52 | 0.78 |
ICSBP | 606 | + | 0.81 | 0.69 |
MYB | 601 | - | 0.76 | 0.77 |
Pax-2 | 604 | - | 0.76 | 0.58 |
SMAD | 603 | + | 0.73 | 0.73 |
Functional Effect of ICSBP/IRF-8 in the TNFR1_{-609(C/T)} SNP
On the other hand, we found ICSBP (also known as IRF-8) to be directly related with the purpose of our study. ICSBP/IRF-8 shows a preference for binding the T allele (see Table 5). As a member of IRF family of transcription factors it is an important modulator of IFNγ signalling cascade and was identified in association on the promoter region of numerous macrophage essential genes such as IL12, IL1β, IL18, iNOS or ISG15 [39].
In addition, several genes regulated by ICSBP/IRF-8, such as MAP4K4, IL-17R, and SOCS7, are involved in different stages of the nuclear factor κ B (NFκ B) signaling pathway [39]. Therefore, we can hypothesize that ICSBP/IRF-8 transcription factor might be also regulating the NFκ B signaling pathway through the control of the first gene of this signalling cascade, the TNFR1 gene. In support of this hypothesis, Zhao et al. established that ICSBP/IRF-8 and TNFR1 are closely related genes [40]. They found ICSBP/IRF-8 to be associated with an enhanced ubiquination of TNFR associated factor 6 (TRAF6), a protein that mediate the signal transduction from members of the TNF receptor superfamily, and the activation of AP-1 and NFκ B transcription factors.
On the other hand, several studies demonstrated that ICSBP/IRF-8 promotes the differentiation and activation of dendritic cells and macrophages cells [41, 42], and that, at the same time, TNFR1 mRNA level is increased during this biological process [43].
Discussion
We have introduced a new IFS-based approach for scoring DNA sequences against DNA motifs called SC_{ intuit }. In this work we review three scoring schemes. These approaches have several drawbacks. SC_{ indep } is based on an incorrect assumption that the nucleotides of a given TFBS are independent. In that context, SC_{ dep } extended the score in order to account for positional dependencies. The problems associated with unnormalized scores at each position have been pointed out [15]. In addition, the results vary depending on the choice of the method and parameters for testing the dependencies. The main drawback with SC_{ mat } is that it has a tendency to overlearn the training data and consequently its performance decreases when applied to real problems. There is therefore a need for a scoring method that accounts for positional dependencies without compromising either the consistency or the accuracy of the results.
As explained above, SC_{ intuit } is based on the IFS theory, which has been successfully applied to problems that suffers from noisy and imprecise data. IFS theory represents uncertainty with respect to both the degree of membership and non-membership. The uncertainty associated with the tasks of scoring DNA sequences against motifs makes intuitionistic concepts particularly suitable for handling this kind of data. Taking advantage of such properties, we define the membership and non-membership degrees of a given pair bases at a given position not only as a function of their combined probability of occurrence, but also taking into account the importance of each individual base at its corresponding position.
One of the biggest issues for this kind of scoring methods is giving high scores for the known binding sequences of the motifs without overfitting. Our proposed approach adequately solves the problem of computing the score of a given sequence against a given motif by considering the binding sequences that comprise the motif not only individually but also as part of such set of sequences. Simple experiments shows how other methods fail in capturing realistic differences, while SC_{ intuit } provides good results (Figure 1, 2). Our method assigned high scores for known binding sites, disfavouring mutations in the conserved positions of the binding site.
These insights are confirmed from experiments for predicting TFBSs in large datasets. We compared the performance of the proposed scoring methods on recognizing motifs in sets of random sequences from a third-order Markov model background distribution in two circumstances: i) when inserting known binding sequences, and ii) when inserting mutated binding sequences. In both situations we found that our proposed method gave the smallest number of false positives per TF whilst simultaneously giving a high number of true positives (Figures 3, 4). More importantly, our method outperforms the other approaches when dealing with real experimental data derived from Chip-seq assays. In this case, again, the number of false positive is significantly reduced (Figure 5). Finally, we validated our method studying the effect of known SNPs of TNFR1 gene in the binding affinity of TRANSFAC TFs for the response against Aspergillus fumigatus. We found the highest scoring for ICSBP TF among all the human TRANSFAC motifs. Although functional analysis should be performed, according to several previous studies, we hypothesize that the presence of TNFR1_{-609(G/T)}polymorphisms could be used to predict susceptibility to infection of hematological patients.
In general, the obtained results on the different experiments demonstrated that the proposed intuitionistic approach provide a better and more accurate model for the detection of motifs and for the relationships between positions of the TFBSs.
Conclusions
In the present study, we have introduced SC_{ intuit }, a new scoring method for measuring sequence-motif affinity, based on IFS theory. Our main objective was to improve the prediction quality for TFs of the existing approaches, reducing the false positive rate without compromising sensitivity. We show that SC_{ intuit } outperforms other approaches in motif recognition tasks, and prove how it can be successfully applied to real research problems. We have used our approach as a scanning method for the prediction of TFBSs, but it also can be incorporated with methods for de novo discovery of motifs. As intuitionistic theory is specially suitable for problems that deal with imprecise concepts, we are currently working on a fuzzy approach that applies the proposed scoring in an ab initio method to find motifs in large sets of related DNA sequences.
Methods
Alternative approaches
In recent years, several scoring methods for the prediction of TFBSs have been proposed. In this section we give a brief overview of those methods that take account of positional dependencies, as they have been shown to outperform methods that assume independence. Let us first introduce the notation. Let B = {A, C, G, T} be the set of the four DNA nucleotides. Let D be a set of ordered DNA sequences on B of length n. Let us suppose that we have a motif M = S_{1},..., S_{ t }, where S_{ i } is a DNA sequence on D consisting of t aligned binding sites of length n. The problem is then reduced to assigning a score to the pair formed by a given putative TFBS, S ∈ D, and a given motif, M.
In what follows we will follow the notation proposed by Wasserman and Sandelin in [44], where F (b, i), for b ∈ B and 1 ≤ i ≤ n shows the occurrences of nucleotide b in position i, and $P(b,i)=\frac{F(b,i)}{t}+a(b)$, for b∈ B and 1 ≤ i ≤ n is the corrected probability of base b at position i, where a(b) is a smoothing parameter (a(b) = 0.001). a(b) = 0.01 is usually reported but our experiments show that smaller values provide more accurate results.
Statistical dependencies
The first step for extending this score involves testing the dependencies between each pair of positions i and j. The authors introduced three different methods: i) χ^{2} test; ii) G statistics; and iii) Bayesian hypothesis testing. The authors used these three methods to calculate the dependencies between pairs of positions in the motifs available in the public database JASPAR [23]. The reader should note that the accurate computation of positional dependencies is still an open problem since different results are obtained depending on the method and parameters used in their computation (see Supplementary Material 2-4 in [14]). Further details about obtaining the position dependencies and multiple test corrections can be found in [14].
where a(b_{1},..., b_{ m }) = a(b_{1}) ... a(b_{ m }) is a smoothing parameter.
where, k_{1} is the number of independent positions, k_{2} is the number of dependent positions of order 2 (nucleotides at positions j_{ i }and j_{i+1}) and k_{ m }the number of dependent positions of order m (nucleotides at positions j_{ i }, j_{i+1},..., j_{i+m-1}).
Matrix based
where b_{1,} b_{2} ∈ 2 B and 1 ≤ i, j ≤ n and i ≠ j. For more on this method see [15].
In order to obtain a normalized value for the score, equation (6) should be applied.
Intuitionistic fuzzy sets
Intuitionistic fuzzy sets (IFS) theory was proposed by Atanassov [17]. It has been applied in such diverse fields as decision making [45], logic programming [46] medical diagnosis [47, 48], pattern recognition [49], etc. IFS theory is an extension of the fuzzy sets theory previously proposed by Zadeh [16] that allows the degrees of membership and non-membership to be independently uncertain, which makes the representation more flexible at capturing the current state of our understanding given inconclusive data [50, 51]. Next, we introduce some basic IFS concepts.
where μ_{ A }, ν_{ A }: X → 0[1] denote membership function and non-membership function of A, satisfying 0 ≤ μ_{ A } + ν_{ A }≤ 1 for every x ∈ X. Therefore, the degree of uncertainty of x to A is π_{ A }(x) = 1 - μ_{ A } - ν_{ A }. For more on this topic please refer to [17, 50, 51].
Intuitionistic representation of motifs
where B × B is the universe of discourse, i.e. the set of all 16 possible combinations of bases for two given positions i and j (AA, AC,..., TT).
Membership degree computation
where the above notation holds. As can be seen, the membership degree is a function of the probability of the pair of bases being compared and their individual conservation. Obviously, $0\le {\mu}_{{I}_{i,j}^{M}}({b}_{1},\phantom{\rule{0.1em}{0ex}}{b}_{2})\le 1$ and the degree increases as do the corrected probabilities of bases b_{1} and b_{2} in positions i and j, as well as the individual corrected probabilities P (b_{1,} i) and P (b_{2,} j).
Non-membership degree computation
where $I{C}_{p}^{b}=\frac{2+P(b,p){\mathrm{log}}_{2}(P(b,p))}{2}$ is the normalized information content of base b in position p and ${\nu}_{{I}_{i,j}^{M}}({b}_{1},\phantom{\rule{0.1em}{0ex}}{b}_{2})$ is in the range $0\le {\nu}_{{I}_{i,j}^{M}}({b}_{1},\phantom{\rule{0.1em}{0ex}}{b}_{2})\le 1$. The reader should note that we make use of the IC of the two positions i, j to assign the allowed degree of uncertainty for such pair of positions. When the IC is high, the degree of uncertainty will be low and viceversa. Likewise, as the information content of the two basis increases, the sum of the membership degrees, ${\mu}_{{I}_{i,j}^{M}}({b}_{1},\phantom{\rule{0.1em}{0ex}}{b}_{2})+{\nu}_{{I}_{i,j}^{M}}({b}_{1},\phantom{\rule{0.1em}{0ex}}{b}_{2})$, gets closer to 1. Therefore, the non-membership degree in positions i, j is a function of the corresponding membership degree and the uncertainty level computed for b_{1} and b_{2}. It is easy to prove that ${\mu}_{{I}_{i,j}^{M}}({b}_{1},\phantom{\rule{0.1em}{0ex}}{b}_{2})+{\nu}_{{I}_{i,j}^{M}}({b}_{1},\phantom{\rule{0.1em}{0ex}}{b}_{2})\le 1$
Scoring
where $\mathrm{max}({\nu}_{{I}_{i,j}^{M}})$ is the maximum degree of non-membership in M found in the pair of positions i and j considering all the possible combination of basis b_{1}, b_{2} ∈ B^{2}, and ${\mu}_{{I}_{i,j}^{M}}({b}_{1},\phantom{\rule{0.1em}{0ex}}{b}_{2})$ and ${\nu}_{{I}_{i,j}^{M}}({b}_{1},\phantom{\rule{0.1em}{0ex}}{b}_{2})$ are the membership degree and non-membership degree of the pairs for the basis (b_{1}, b_{2}) ∈ B in the pair of positions i, j of M, computed as stated in sections and respectively.
where are the $\mathrm{min}(S{C}_{intuit}^{i,j})$ and $\text{max}(S{C}_{intuit}^{i,j})$ are the min/max possible scores in the positions (i, j) of the motif.
Declarations
Acknowledgements
We would like to thank the anonymous reviewers for their helpful comments which significantly improved the quality of the paper. This work has been carried out as part of projects P08-TIC-4299 of J. A., Sevilla and TIN2009-13489 of DGICT, Madrid.
Authors’ Affiliations
References
- Bailey T, Elkan C: Fitting a mixture model by expectation maximization to discover motifs in biopolymers. Proc Int Conf Intell Syst Mol Biol 1994, 2: 28–36. Citeseer CiteseerPubMedGoogle Scholar
- Lawrence C, Altschul S, Boguski M, Liu J, Neuwald A, Wootton J: Detecting subtle sequence signals: a Gibbs sampling strategy for multiple alignment. Chem Rev 1993, 93: 741. 10.1021/cr00018a006View ArticleGoogle Scholar
- Hughes J, Estep P, Tavazoie S, Church G: Computational identification of cis-regulatory elements associated with groups of functionally related genes in Saccharomyces cerevisiae. Journal of molecular biology 2000, 296(5):1205–1214. 10.1006/jmbi.2000.3519View ArticlePubMedGoogle Scholar
- Gordan R, Narlikar L, Hartemink A: Finding regulatory DNA motifs using alignment-free evolutionary conservation information. Nucleic Acids Research 2010, 38(6):e90. 10.1093/nar/gkp1166View ArticlePubMedPubMed CentralGoogle Scholar
- Haudry Y, Ramialison M, Paten B, Wittbrodt J, Ettwiller L: Using Trawler standalone to discover overrepresented motifs in DNA and RNA sequences derived from various experiments including chromatin immunoprecipitation. Nature Protocols 2010, 5(2):323–334. 10.1038/nprot.2009.158View ArticlePubMedGoogle Scholar
- Das M, Dai H: A survey of DNA motif finding algorithms. BMC Bioinformatics 2007, 8(Suppl 7):S21. 10.1186/1471-2105-8-S7-S21View ArticlePubMedPubMed CentralGoogle Scholar
- Hertz G, Hartzell G III, Stormo G: Identification of consensus patterns in unaligned DNA sequences known to be functionally related. Bioinformatics 1990, 6(2):81. 10.1093/bioinformatics/6.2.81View ArticleGoogle Scholar
- Sandelin A, Wasserman W, Lenhard B: ConSite: web-based prediction of regulatory elements using cross-species comparison. Nucleic acids research 2004, (32 Web Server):W249. 10.1093/nar/gkh372Google Scholar
- Benos P, Lapedes A, Stormo G: Probabilistic code for DNA recognition by proteins of the EGR family. Journal of molecular biology 2002, 323(4):701–727. 10.1016/S0022-2836(02)00917-8View ArticlePubMedGoogle Scholar
- Bulyk M, Johnson P, Church G: Nucleotides of transcription factor binding sites exert interdependent effects on the binding affinities of transcription factors. Nucleic acids research 2002, 30(5):1255. 10.1093/nar/30.5.1255View ArticlePubMedPubMed CentralGoogle Scholar
- Eisen M: All motifs are not created equal: structural properties of transcription factor-DNA interactions and the inference of sequences specificity. Genome Biology 2005, 6(5):P7. 10.1186/gb-2005-6-5-p7View ArticleGoogle Scholar
- Won K, Sandelin A, Marstrand T, Krogh A: Modeling promoter grammars with evolving hidden Markov models. Bioinformatics 2008, 24(15):1669. 10.1093/bioinformatics/btn254View ArticlePubMedGoogle Scholar
- Kim N, Tharakaraman K, Mariño-Ramírez L, Spouge J: Finding sequence motifs with Bayesian models incorporating positional information: an application to transcription factor binding sites. BMC bioinformatics 2008, 9: 262. 10.1186/1471-2105-9-262View ArticlePubMedPubMed CentralGoogle Scholar
- Tomovic A, Oakeley E: Position dependencies in transcription factor binding sites. Bioinformatics 2007, 23(8):933. 10.1093/bioinformatics/btm055View ArticlePubMedGoogle Scholar
- Zare-Mirakabad F, Ahrabian H, Sadeghi M, Nowzari-Dalini A, Goliaei B: New scoring schema for finding motifs in DNA Sequences. BMC Bioinformatics 2009, 10: 93. 10.1186/1471-2105-10-93View ArticlePubMedPubMed CentralGoogle Scholar
- Zadeh L: Fuzzy sets*. Information and control 1965, 8(3):338–353. 10.1016/S0019-9958(65)90241-XView ArticleGoogle Scholar
- Atanassov K: Intuitionistic fuzzy sets. Fuzzy Sets and Systems 1986, 20: 87–96. 10.1016/S0165-0114(86)80034-3View ArticleGoogle Scholar
- Garcia F, Lopez F, Cano C, Blanco A: FISim: A new similarity measure between transcription factor binding sites based on the fuzzy integral. BMC Bioinformatics 2009, 10: 224. 10.1186/1471-2105-10-224View ArticlePubMedPubMed CentralGoogle Scholar
- Lopez F, Blanco A, Garcia F, Cano C, Marin A: Fuzzy association rules for biological data analysis: a case study on yeast. BMC Bioinformatics 2008, 9: 107. 10.1186/1471-2105-9-107View ArticlePubMedPubMed CentralGoogle Scholar
- Liang L, Mandal V, Lu Y, Kumar D: MCM-test: a fuzzy-set-theory-based approach to differential analysis of gene pathways. BMC Bioinformatics 2008, 9(6):S16. 10.1186/1471-2105-9-S6-S16View ArticlePubMedPubMed CentralGoogle Scholar
- Chen J, Bardes E, Aronow B, Jegga A: ToppGene Suite for gene list enrichment analysis and candidate gene prioritization. Nucleic Acids Research 2009, (37 Web Server):W305. 10.1093/nar/gkp427Google Scholar
- Lam F, Steger D, O'Shea E: Chromatin decouples promoter threshold from dynamic range. Nature 2008, 453(7192):246–250. 10.1038/nature06867View ArticlePubMedPubMed CentralGoogle Scholar
- Sandelin A, Alkema W, Engstrom P, Wasserman W, Lenhard B: JASPAR: an open-access database for eukaryotic transcription factor binding profiles. Nucleic Acids Research 2004, (32 Database):D91. 10.1093/nar/gkh012Google Scholar
- Van Helden J: Regulatory sequence analysis tools. Nucleic acids research 2003, 31(13):3593. 10.1093/nar/gkg567View ArticlePubMedPubMed CentralGoogle Scholar
- Hanley J, McNeil B: The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology 1982, 143: 29–36.View ArticlePubMedGoogle Scholar
- Fawcett T: An introduction to ROC analysis. Pattern recognition letters 2006, 27(8):861–874. 10.1016/j.patrec.2005.10.010View ArticleGoogle Scholar
- Chen X, Xu H, Yuan P, Fang F, Huss M, Vega V, Wong E, Orlov Y, Zhang W, Jiang J, et al.: Integration of external signaling pathways with the core transcriptional network in embryonic stem cells. Cell 2008, 133(6):1106–1117. 10.1016/j.cell.2008.04.043View ArticlePubMedGoogle Scholar
- Sharov A, Ko M: Exhaustive Search for Over-represented DNA Sequence Motifs with CisFinder. DNA Research 2009, 16:(5):261–73. 10.1093/dnares/dsp014View ArticlePubMedPubMed CentralGoogle Scholar
- Matys V, Kel-Margoulis O, Fricke E, Liebich I, Land S, Barre-Dirrie A, Reuter I, Chekmenev D, Krull M, Hornischer K, et al.: TRANSFAC (R) and its module TRANSCompel (R): transcriptional gene regulation in eukaryotes. Nucleic acids research 2006, (34 Database):D108. 10.1093/nar/gkj143Google Scholar
- Denning D: Invasive aspergillosis. Clinical infectious diseases 1998, 781–803. 10.1086/513943Google Scholar
- Offner F, Cordonnier C, Ljungman P, Prentice H, Engelhard D, Bacquer D, Pauw B: Impact of previous aspergillosis on the outcome of bone marrow transplantation. Clinical infectious diseases 1998, 26(5):1098–1103. 10.1086/520274View ArticlePubMedGoogle Scholar
- Chamilos G, Luna M, Lewis R, Bodey G, Chemaly R, Tarrand J, Safdar A, Raad I, Kontoyiannis D: Invasive fungal infections in patients with hematologic malignancies in a tertiary care cancer center: an autopsy study over a 15-year period (1989–2003). Haematologica 2006, 91(7):986.PubMedGoogle Scholar
- Diop G, Spadoni J, Do H, Hirtzig T, Coulonges C, Labib T, Issing W, Rappaport J, Therwath A, Lathrop M, et al.: Genomic approach of AIDS pathogenesis: exhaustive genotyping of the TNFR1 gene in a French AIDS cohort. Biomedicine & Pharmacotherapy 2005, 59(8):474–480.View ArticleGoogle Scholar
- Hehlgans T, Pfeffer K: The intriguing biology of the tumour necrosis factor/tumour necrosis factor receptor superfamily: players, rules and the games. Immunology 2005, 115: 1. 10.1111/j.1365-2567.2005.02143.xView ArticlePubMedPubMed CentralGoogle Scholar
- Bochud P, Chien J, Marr K, Leisenring W, Upton A, Janer M, Rodrigues S, Li S, Hansen J, Zhao L, et al.: Toll-like receptor 4 polymorphisms and aspergillosis in stem-cell transplantation. New England Journal of Medicine 2008, 359(17):1766. 10.1056/NEJMoa0802629View ArticlePubMedPubMed CentralGoogle Scholar
- Baker E, Chen L, Smith C, Callen D, Goodwin R, Sutherland G: Chromosomal location of the human tumor necrosis factor receptor genes. Cytogenetics and cell genetics 1991, 57:(2–3):117.View ArticleGoogle Scholar
- Sainz J, Salas I, Lopez E, Olmedo C, Comino A, Garcia F, Blanco A, Oyonarte S, Bueno P, Jurado M: TNFR1 mRNA Expression Level and TNFR1 Gene Polymorphisms Are Predictive Markers for Susceptibility to Develop Invasive Pulmonary Aspergillosis. International Journal of Immunopathology and Pharmacology 2009, 22(3):557–565.Google Scholar
- Wingender E: TheTRANSFAC project as an example of framework technology that supports the analysis of genomic regulation. Brief Bioinform 2008, 9: 326–332. 10.1093/bib/bbn016View ArticlePubMedGoogle Scholar
- Dror N, Alter-Koltunoff M, Azriel A, Amariglio N, Jacob-Hirsch J, Zeligson S, Morgenstern A, Tamura T, Hauser H, Rechavi G, et al.: Identification of IRF-8 and IRF-1 target genes in activated macrophages. Molecular immunology 2007, 44(4):338–346. 10.1016/j.molimm.2006.02.026View ArticlePubMedGoogle Scholar
- Zhao J, Kong H, Li H, Huang B, Yang M, Zhu C, Bogunovic M, Zheng F, Mayer L, Ozato K, et al.: IRF-8/interferon (IFN) consensus sequence-binding protein is involved in Toll-like receptor (TLR) signaling and contributes to the cross-talk between TLR and IFN-γ signaling pathways. Journal of Biological Chemistry 2006, 281(15):10073. 10.1074/jbc.M507788200View ArticlePubMedGoogle Scholar
- Tamura T, Ozato K: Review: ICSBP/IRF-8: its regulatory roles in the development of myeloid cells. Journal of Interferon & Cytokine Research 2002, 22: 145–152.View ArticleGoogle Scholar
- Tamura T, Nagamura-Inoue T, Shmeltzer Z, Kuwata T, Ozato K: ICSBP directs bipotential myeloid progenitor cells to differentiate into mature macrophages. Immunity 2000, 13(2):155–165. 10.1016/S1074-7613(00)00016-9View ArticlePubMedGoogle Scholar
- Schling P, Rudolph C, Heimerl S, Fruth S, Schmitz G: Expression of tumor necrosis factor alpha and its receptors during cellular differentiation. Cytokine 2006, 33(5):239–245. 10.1016/j.cyto.2006.02.007View ArticlePubMedGoogle Scholar
- Wasserman W, Sandelin A: Applied bioinformatics for the identification of regulatory elements. Nature Reviews Genetics 2004, 5(4):276–287. 10.1038/nrg1315View ArticlePubMedGoogle Scholar
- Szmidt E, Kacprzyk J: Intuitionistic fuzzy sets in group decision making. Notes on IFS 1996, 2: 11–14.Google Scholar
- Atanassov K, Gargov G: Intuitionistic fuzzy logic. Compt Rend Acad Bulg Sci 1990, 43: 9–12.Google Scholar
- De S, Biswas R, Roy A: An application of intuitionistic fuzzy sets in medical diagnosis. Fuzzy Sets and Systems 2001, 117(2):209–213. 10.1016/S0165-0114(98)00235-8View ArticleGoogle Scholar
- Khatibi V, Montazer G: Intuitionistic fuzzy set vs. fuzzy set application in medical pattern recognition. Artificial Intelligence In Medicine 2009, 47: 43–52. 10.1016/j.artmed.2009.03.002View ArticlePubMedGoogle Scholar
- Hung W, Yang M: Similarity measures of intuitionistic fuzzy sets based on Hausdorff distance. Pattern Recognition Letters 2004, 25(14):1603–1611. 10.1016/j.patrec.2004.06.006View ArticleGoogle Scholar
- Atanassov K: New operations defined over the intuitionistic fuzzy sets. Fuzzy Sets and Systems 1994, 61(2):142. 10.1016/0165-0114(94)90229-1View ArticleGoogle Scholar
- Atanassov K: Intuitionistic Fuzzy Sets: Theory and Applications. Physica-Verlag, Heidelberg, New York 1999.Google 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.