Estimating developmental states of tumors and normal tissues using a linear time-ordered model
- Bo Zhang†1, 2,
- Beibei Chen†1, 2,
- Tao Wu1,
- Zhenyu Xuan3,
- Xiaopeng Zhu1Email author and
- Runsheng Chen1Email author
© Zhang et al; licensee BioMed Central Ltd. 2011
Received: 25 June 2010
Accepted: 11 February 2011
Published: 11 February 2011
Tumor cells are considered to have an aberrant cell state, and some evidence indicates different development states appearing in the tumorigenesis. Embryonic development and stem cell differentiation are ordered processes in which the sequence of events over time is highly conserved. The "cancer attractor" concept integrates normal developmental processes and tumorigenesis into a high-dimensional "cell state space", and provides a reasonable explanation of the relationship between these two biological processes from theoretical viewpoint. However, it is hard to describe such relationship by using existed experimental data; moreover, the measurement of different development states is also difficult.
Here, by applying a novel time-ordered linear model based on a co-bisector which represents the joint direction of a series of vectors, we described the trajectories of development process by a line and showed different developmental states of tumor cells from developmental timescale perspective in a cell state space. This model was used to transform time-course developmental expression profiles of human ESCs, normal mouse liver, ovary and lung tissue into "cell developmental state lines". Then these cell state lines were applied to observe the developmental states of different tumors and their corresponding normal samples. Mouse liver and ovarian tumors showed different similarity to early development stage. Similarly, human glioma cells and ovarian tumors became developmentally "younger".
The time-ordered linear model captured linear projected development trajectories in a cell state space. Meanwhile it also reflected the change tendency of gene expression over time from the developmental timescale perspective, and our finding indicated different development states during tumorigenesis processes in different tissues.
Cancer is a severe threat to human health. Although there are many established methods for overcoming this disease, the high mortality caused by cancer is still a severe threat to human. Meanwhile, the side-effects of many therapeutic methods greatly affect the quality of life of individuals and their families. Uncertainty about the mechanisms of tumorigenesis greatly handicaps the creation and application of suitable therapeutic methods. Tumorigenesis is a complex process, affected by both genetic factors and environmental conditions. There is evidence to suggest that developmental processes and tumorigenesis share some conserved mechanisms [1, 2]. Time-course microarray experiments have the advantage of allowing us to study the dynamics of gene regulation. Time-course microarrays have recently been used to identify biological markers associated with disease and to examine the expression patterns of genes that are important in tumorigenesis and development [1, 3].
Many models have been proposed to explain the process of tumorigenesis and its relationship to development. The "cancer attractor" model was first suggested by Kauffman in the 1970 s  and can be used to explain how a Gene Regulation Network (GRN) confers a single genome with the capacity to produce a diversity of stable, discretely distinct cell types over the process of development . Foster introduced a simplified differential equation described by Huang  into a model containing two genes. Five hundred "cells" were stimulated to "differentiate", finally reaching the "stable attractors" position, demonstrating the validity of the "cancer attractor" model. There is a significant amount of evidence based on time-course microarray experiments which supports the attractor theory [5, 7–9]. Mar and Quackenbush  have recently decomposed cell fate transition into two processes: the core process that includes the main differentiation pathway, and a transient process that captures information from the environment and controls the core process.
Cell state space is a high-dimensional space in which different cell types correspond to points or distributions . In Foster's work  a system based on two genes generated 3-dimensional coordinates including two gene dimensions and one "quasi potential" dimension, however, that still exists some difficulties to explain the biological meaning of this "quasi potential" dimension.. Since time is invariable and irreversible, sequentially ordered developmental progression is a very important innate characteristic of life. If we treat time as a scale for measuring cell state space, it is possible to describe the high-dimensional cell state space by a low-dimensional space.
Many approaches, including PCA and SVD methods [12–14], the Bayesian models , HMM(Hidden Markov Models) , and some ANOVA and regression-based model [17, 18] have been applied for the analysis of time-course microarray data from different aspects. Most of these methods are designed to detect genes which undergo significant changes and to classify expression patterns in time-course experiments. Only few methods emphasize temporal order within experiments and time-course expression profiles.
Here, in order to capture the temporal properties and describe the trajectories of development processes, we propose a new linear model, named the "time-ordered linear model", which draws on the idea that a co-bisector can represent the main tendency of a series of vectors. This co-bisector model has two main advantages: first, unlike present methods such as PCA, the biological meaning of the co-bisector model is borne in mind in the design of the model. A co-bisector conserves the temporal properties of a series of vectors since they have order-restricted projection locations on the co-bisector. Furthermore, our model preserves the spatial distance ratio between neighboring samples which have fixed locations in microarray space. Our time-ordered linear model can be used as a measurement scale of gene expression variation in microarray space, thus creating a new application for time-course microarray data; estimating the expression pattern similarities between expression data from more than one source. In the present work, we apply our time-ordered linear model to estimate expression pattern similarities between different tumor tissues and their corresponding normal tissues in both mice and humans. Our time-ordered linear model describe the trajectories of development process in a cell state space from the gene expression pattern perspective, thus helping us to improve our understanding of the relationships among different cell types in cell state space.
Design the Time-ordered Linear Model in the Abstract Cell State Space
The concept of cell state space was proposed by Kauffman. In high-dimensional cell state space, cell types with similar properties are grouped together. The dimensions of cell state space are measurements of cell properties such as SNP, transcriptome, and epigenetic modification. The expression pattern of a cell is simply a reflection of its cell state. Recently, the mouse and human genome DNA methylation maps have been reported [19, 20]. We believe that a fuller and more detailed description of cell state space will emerge as more and more high-throughput data are published. But the work presented here only focus on gene expression patterns, and we simplified the cell state space as a microarray space which dimensions are determined by genes.
In cell differentiation microarray experiments, samples representing different cell differentiation stages i.e. different time points, are linked in order by a curve in microarray space. As a powerful approach, the PCA method can easily draw the mathematical distribution of principal components of these points to maximize the sum of variance. In order to test the efficiency of description development trajectory by PCA methods, we analyzed dataset GSE13149 which represents mouse fetal liver development. Although some principal components (PCs) may preserve the sampling order (Figure S1), one hand, the biological meaning of these PCs can not be derived directly from the PCA methods. Some approaches, such as GO annotation, were applied to obtain the biological meaning of these PCs from the biological function of significantly changed genes. On the other hand, under the condition that maximizes sum of variance, the projection distance ratio of neighbor points on PCs are different from real distance ratio in microarray space (Table S1). For these two reasons, it is difficult to use the PCA method to describe the trajectories of the development process.
Using this time-ordered linear model we can obtain "cell developmental state lines" representing the temporal properties of differentiation trajectories. The main advantages of this time-ordered model are that the sequential order is preserved and that the distance between points is maximized, ensuring that developmental processes are in the right order and that the relationship among neighboring points in microarray space is denoted accurately. Here, by analyzing published tissue development and cell differentiation expression profiles obtained using time-course experiments (described below), we obtained cell developmental state lines representing several developmental and differentiation processes, and then compared published tumor expression profiles obtained using the same microarray platform (described below). By calculating the projection positions of expression profiles on the cell developmental state line, and their relationship, we were able to deduce developmental states of these tumors in these processes on a developmental-temporal scale.
Similar to PCA, Time-ordered linear model generate one principal component to represent a mass of sample points. This principal component which we called 'developmental state line', is not only representing one mathematical characteristic of samples, but also reflects temporal property of a development process. In other words, the biological meaning was denoted to a mathematical characteristic, and this clear biological development perspective enable us to accurately estimate different development stage of same tissues samples came from different sources by one line.
Distance and Distance Ratio in microarray Space, developmental state line (DSL) and PCs of dataset GSE5334
Distance in microarray space (i->i+1)
Distance ratio in microarray space ((i->i+1): (i+1->i+2))
Distance on DSL (i->i+1)
Distance ratio on DSL ((i->i+1): (i+1->i+2))
Distance on PC1 (i->i+1)
Distance ratio on PC1 ((i->i+1): (i+1->i+2))
Distance on PC2 (i->i+1)
Distance ratio on PC2 ((i->i+1): (i+1->i+2))
The robustness of this time-ordered linear model was also tested. We generated 3 cell developmental state lines by removing 1, 2 and 3 points from dataset GSE13149. Then dataset GSE6998 were projected to these 3 cell developmental state lines one by one. The results indicated that the projection locations of the test dataset GSE6998 maintained the order following interval of points of the cell development state lines. Meanwhile, the mean values of the projection locations were similar to the control, and the variance constantly increased. Especially the first test point 10.5 D and last test point 16.5 D, suffered bigger variance than other points (Table S2 and Figure S2). Interestingly, lack of the first two or three points influenced the resolution of later development period samples. This result indicated that the model has a high robustness, especially for the lack of medial samples, but lack of samples at either end of the time points would influence the projection result.
Mouse liver cell developmental state line demonstrates apparent "similarity to early developmental stage" of liver tumors
We projected the dataset GSE 5128 which contains a series of carcinogen-treated samples onto the liver cell developmental state line. Interestingly, a dynamic back-moving tendency appeared (Figure 6). Mice treated with the two positive liver carcinogens had the earliest projection positions (those treated with 1,5-Naphthalenediamine located at46.52 ± 0.54, while those treated with 2,3-Benzofuran located at 47.70 ± 1.40). Projection positions for mice treated with the two negative liver carcinogens located in the middle between control and positive liver carcinogen-treated mice. These results suggest that there is a link between carcinogenicity and the apparent "younger developmental state" of cells observed here; different carcinogens change the cell state to different degrees. Our liver cell developmental state line represents the changes in liver cell state associated with development. The analysis of these two datasets suggested a fact that liver tumors have a similar cell state to that of earlier developmental stages of the fetal liver. As is well known, tumor or cancer cells have many of the properties of self-renewing stem cells. It is accepted that once the cell status of cancer cells departs from that of normal cells, they regain the ability to proliferate uncontrollably.
Mouse ovary tumor demonstrates apparent "developmentally younger"
We conducted a similar analysis of ovary development time-course experimental data. In dataset GSE 5334, expression profiles from 5 stages (Gestational days GD11 day to GD18) were used to construct the ovary cell developmental state line as described above. Dataset GSE6916 was used to test the accuracy of the cell developmental state line. The results indicated that the cell developmental state line ordered each developmental stage accurately (Figure 4). We then used the cell developmental state line to estimate the projection positions of dataset GSE5987, which contains 7 ovarian tumors and 4 normal ovary samples. Results were also similar to those described above for the liver (Figure 4), once again suggesting that the process of tumorigenesis in ovary make ovary tumor cells have a high "similarity to early developmental stage".
Cell state variation caused by carcinogenesis in the lung do no share same direction to mouse lung cell developmental state line
Three sets of expression data from different developmental stages were projected onto the lung cell developmental state line (based on dataset GSE11539 which represents lung development from embryonic day 11.5 to postnatal day 5) (Figure S3). The projection positions of each of the samples were distributed from early to terminal differentiated stages according to the age of the mice. This result demonstrates that the cell developmental state line can faithfully represent the temporal properties of development. Then the cell developmental state line was used to estimate the states of GSE5127 lung carcinogenesis expression dataset, in which different carcinogens were feed to the mouse. The results were different to those for the liver and ovary. The four chemicals hardly changed the "developmental state" of lung cell state (Figure S4). Although 2,3-Benzofuran and 1,5-Naphthalenediamine can cause cancer in both liver and lung, our model indicates that, different to in the liver and ovary, the genes belonged to a developmental pathway that may have not been involved in carcinogenesis process in the lung.
Generally speaking, the cell developmental state lines generated by our time-ordered linear model using time-course microarray experimental data accurately reflect the gene expression pattern variation over time during development, and their utility for estimating the relationship between tumor cell state and normal cell state could supply clues for further investigations of the tumorigenesis mechanism. If tissue-specific tumors are treated as "attractors", the cell developmental state line describes the relative position of the "attractors" in cell state space. In the liver and the ovary, cancer attractors may be located nearer to earlier developmental stages than is the case for normal tissues. In the lung, cancer attractors may be located in a direction that is vertical to the developmental direction, and the cell developmental state line can not distinguish such cell state changes. Projection results for cancer samples give an indication that the mechanisms of tumorigenesis may be not the same in different tissues.
Mouse tissues development trajectories have different directions in cell state space
As known, the liver and lung both come from the endoderm, and the ovary is developed from mesoderm. We guess that if the origins of tissues in the gastrula are same, their development trajectories may share more commons in the cell state space, and the angle between these tissue developmental state lines would be smaller.
Cell differentiation state lines of human ES cell distinguished cell states variation of human tumors
Since human embryonic stem cells are an important model for studying human development, we transformed expression data from ES cell differentiation time-course experiments to a cell developmental state line representing the ES cell differentiation process. Unlike mouse tissues developments start from different cell states, we calculated two cell developmental state lines for all ES cell differentiation processes starting from the same pluripotent cell state. This generated 2-dimensional coordinates, with each cell developmental state line axis representing different developmental processes. We used the GSE9940 dataset of the ES cell-derived neural rosette differentiation expression profile to generate a "neuronal" cell developmental state line, and the GSE8884 dataset of the ES cell-derived blast cell differentiation expression profile to generate a "blast cell" cell developmental state line. We combined two axes to one "differentiation index coordinate", and used it to estimate projection positions of different cancers and normal tissues.
Then the "differentiation-index coordinates" was used to detect cell state of four human tumor expression data (GSE7305, GSE4107, GSE5674, and GSE18520) containing normal tissue samples and diseased tissue samples (endometrium, colon, breast, and ovary, respectively) from the GEO (Gene Expression Omnibus). Moreover, we selected expression data from GSE2109 (obtained from the IGC Expression Project for Oncology (expO)), which contains expression data from more than 2000 tumor samples derived from more then one hundred tissues. Tumor samples originating from the breast, colon, endometrium and ovary were individually estimated by differentiation-index coordinates, and the results were shown according to tissue.
In the case of breast cancer, the tumor and normal breast tissue occupied overlapping positions in the differentiation-index coordinates (Figure S6). The fact that it was not possible to distinguish tumor samples from normal samples by differentiation-index coordinate approach may be due to that tumorigenesis in the breast may not be related to the developmental pathway, or the breast developmental state line may be perpendicular to the "neuronal" and "blast" cell differentiation state lines. Moreover, a recently published comprehensive genome-level catalogue of breast tumors  revealed the complexity of breast tumors. Such complexities at genome level may also influence the accuracy of cell developmental state line.
ES cell differentiation is a good model for studying the process of development. However, successful cell differentiation under in vitro conditions is only possible in a limited number of cell types. The two human ES cells developmental state lines developed here, based on in vitro datasets, represent only two of the possible ES cell differentiation directions, and thus cannot fully represent the complete ES cell differentiation process under in vivo conditions. The work presented here provides an approach forwards understanding the relationships among distinct differentiation directions in cell state space, of which microarray space is only one subspace. The increasing availability of high-throughput proteomic and epigenetic data and other measures of cell properties will make it possible to investigate other dimensions.
Theoretically, each type of differentiated cell has its unique differentiation direction. Even a single cellular activity can be considered to have a unique direction. Many different directions exist in the whole process of tissue development and cell differentiation, and these different directions reflect distinct aspects of cell properties. Here, we selected the changing tendency of gene expression pattern over time to represent the differentiation trajectory. In the embryonic development process, the diversity of cells increases continuously with cell differentiation. However, it is difficult to describe the relationships among the growing number of cell types, and how the genome facilitates the generation of stable and distinct cell types in the development process is still not clear.
Assuming that differentiation processed follow a strict order, using a time scale should be useful for estimating changes in cell state throughout the differentiation process. The cell developmental state lines generated by time-ordered linear model were able to accurately order different developmental stages in different tissue types. Interestingly, our results did not only suggest that tumorigenesis can be measured by "developmental state lines", but also suggested the possibility all directions of ES cells differentiation can be described. It has been reported that some genes whose expression is significantly altered during tumorigenesis may also play key roles in developmental process [22–24]. Our approach may help to understand the relationship between tumorigenesis and cell differentiation in greater detail.
Tumorigenesis is a complex process. That tumorigenesis shares many similar characteristics with embryonic cell development implies that they have a close, though poorly understood relationship. The "cancer attractor" model presents a new and integrated perspective for viewing these two processes. Constructing cell developmental state lines is an attempt to observe and describe differentiation trajectories in cell state space from another perspective. Our result indicated that even in large scale discrimination, some kinds of tumor showed the same moving tendency at timescale. Although theoretically there are many "cancer attractors" surrounding normal cell state, in realty the number of these "attractors" may be limited to a very small range.
The time-ordered linear model constructed here is an attempt to use a linear trajectory to describe tissue development and cell differentiation processes. Not only it is a novel method to analyze the time-course expression profiles, but by, primarily defining a biological meaning to the mathematical model, the approach also supplies a different viewpoint to traditional methods which mainly emphasize the mathematical characteristics. The time-ordered linear model is a simplified model for calculating lines to represent development trajectories, and some limitations are still need to be addressed. The calculation of this liner model depends on a given order of sample points, and disordered sample points would generate fake cell development state lines. Moreover, in this work, we used the average of several repeats at each time points to calculate the cell development state line. Such an approach partly limits the robustness and increases the sensitivity to noise. In the future, replacing the averaged points by a may overcome these weaknesses.
However, we cannot expect such a simple model to reflect all the details of development, to fit all expression data, or even to distinguish all types of cancer in cell status space. Rather, its function is simply to transform cell differentiation expression profiles to a line in keeping with the natural temporal properties of gene expression during the development processes. The real development trajectories in cell status space are much too complex to be modeled computationally at present. We have therefore started with a simple "line," which captures the basic progression of development from one perspective. If the approach works, it could serve as a basis for further construction of more realistic, comprehensive, and predictive models of cell state space. To draw developmental trajectories in cell state space accurately requires being able to describe cell state from the perspective of all of cell features, including its transcriptome, epigenetic map, and proteome. Understanding the trajectories in cell status space will help reveal the mysteries of embryonic development.
By primarily defining a biological meaning to a mathematical model, we designed the time-ordered linear model which can capture temporal properties of development process, and drew the linearly projected development trajectories in a cell state space. Meanwhile, it reflected the change of gene expression from a developmental timescale perspective. By applying this model to measure tumors of different tissues, we found that different developmental states appeared during tumorigenesis.
Construction of our time-ordered linear model
In N dimensional space, T vectors which are unlooped and head-to-tail jointed have many angle-bisectors when N >> T. In order to unique our linear model, we selected the angle-bisector which maximized the projection of each vectors (Additional file 2), and we named it "co-bisector". Naturally, we selected co-bisector of a series vectors to indicate their main moving tendency. The co-bisector well suited our requirement of a linear model to represent tissue development and cell differentiation processes.
With equation (2)(3)(4)(5), the parameters a i and θ, and vector were determined.
Projecting expression data on the cell developmental state line
We applied our time-ordered linear model to analyze time-course experimental expression profiles obtained from the Gene Expression Omnibus http://www.ncbi.nlm.nih.gov/geo/ for mouse liver, ovary and lung tissue development, and for human ES cell differentiation. We constructed tissue-specific "cell developmental state lines" to represent developmental trajectories and used them to estimate the relationship between tumor tissues and normal tissue development over time. To increase the resolution, we also constructed "differentiation-index coordinates", which consisted of two different "cell developmental state lines" based on the human ES cell differentiation datasets. We used the datasets listed in Table S3 to generate mouse tissue-specific developmental cell developmental state lines. We then used the model to estimate the relationship between the tumor and normal tissue expression profiles listed in Table S4. Human ES cell differentiation cell developmental state lines were generated using the datasets listed in Table S5, and then human cell developmental state line were used to assess the expression profiles of human tumors listed in Table S6.
We thank Dr Geir Skogerbø for greatly useful discussions during model construction and manuscripts revision. We are glad to thank Dr. Lei Xu and Mr. Shikui Tu for kind suggestion. We thank the efforts of IGC for supporting the tumor expression data(expO project).
This research was supported by National Program on Key Basic Research Project (973 Program Grant 2009CB825401), the National Natural Science Foundation of China (60933009) and the Innovation Projects (KSCX2-YW-R-124) from Chinese Academy of Sciences. The founders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.
- Kho AT, Zhao Q, Cai Z, Butte AJ, Kim JY, Pomeroy SL, Rowitch DH, Kohane IS: Conserved mechanisms across development and tumorigenesis revealed by a mouse development perspective of human cancers. Genes Dev 2004, 18(6):629–640. 10.1101/gad.1182504PubMed CentralView ArticlePubMedGoogle Scholar
- Lepourcelet M, Tou L, Cai L, Sawada J, Lazar AJ, Glickman JN, Williamson JA, Everett AD, Redston M, Fox EA, Nakatani Y, Shivdasani RA: Insights into developmental mechanisms and cancers in the mammalian intestine derived from serial analysis of gene expression and study of the hepatoma-derived growth factor (HDGF). Development 2005, 132(2):415–427. 10.1242/dev.01579View ArticlePubMedGoogle Scholar
- Quackenbush J: Microarray analysis and tumor classification. N Engl J Med 2006, 354(23):2463–2472. 10.1056/NEJMra042342View ArticlePubMedGoogle Scholar
- Kauffman S: Differentiation of malignant to benign cells. J Theor Biol 1971, 31(3):429–451. 10.1016/0022-5193(71)90020-8View ArticlePubMedGoogle Scholar
- Foster DV, Foster JG, Huang S, Kauffman SA: A model of sequential branching in hierarchical cell fate determination. J Theor Biol 2009, 260(4):589–597. 10.1016/j.jtbi.2009.07.005View ArticlePubMedGoogle Scholar
- Huang S, Guo YP, May G, Enver T: Bifurcation dynamics in lineage-commitment in bipotent progenitor cells. Dev Biol 2007, 305(2):695–713. 10.1016/j.ydbio.2007.02.036View ArticlePubMedGoogle Scholar
- Huang AC, Hu L, Kauffman SA, Zhang W, Shmulevich I: Using cell fate attractors to uncover transcriptional regulation of HL60 neutrophil differentiation. BMC Syst Biol 2009, 3: 20. 10.1186/1752-0509-3-20PubMed CentralView ArticlePubMedGoogle Scholar
- Chang HH, Hemberg M, Barahona M, Ingber DE, Huang S: Transcriptome-wide noise controls lineage choice in mammalian progenitor cells. Nature 2008, 453(7194):544–547. 10.1038/nature06965View ArticlePubMedGoogle Scholar
- Huang S, Eichler G, Bar-Yam Y, Ingber DE: Cell fates as high-dimensional attractor states of a complex gene regulatory network. Phys Rev Lett 2005, 94(12):128701. 10.1103/PhysRevLett.94.128701View ArticlePubMedGoogle Scholar
- Mar JC, Quackenbush J: Decomposition of gene expression state space trajectories. PLoS Comput Biol 2009, 5(12):e1000626. 10.1371/journal.pcbi.1000626PubMed CentralView ArticlePubMedGoogle Scholar
- Kauffman S: A proposal for using the ensemble approach to understand genetic regulatory networks. J Theor Biol 2004, 230(4):581–590. 10.1016/j.jtbi.2003.12.017View ArticlePubMedGoogle Scholar
- Alter O, Brown PO, Botstein D: Singular value decomposition for genome-wide expression data processing and modeling. Proc Natl Acad Sci USA 2000, 97(18):10101–10106. 10.1073/pnas.97.18.10101PubMed CentralView ArticlePubMedGoogle Scholar
- Holter NS, Mitra M, Maritan A, Cieplak M, Banavar JR, Fedoroff NV: Fundamental patterns underlying gene expression profiles: simplicity from complexity. Proc Natl Acad Sci USA 2000, 97(15):8409–8414. 10.1073/pnas.150242097PubMed CentralView ArticlePubMedGoogle Scholar
- Jonnalagadda S, Srinivasan R: Principal components analysis based methodology to identify differentially expressed genes in time-course microarray data. BMC Bioinformatics 2008, 9: 267. 10.1186/1471-2105-9-267PubMed CentralView ArticlePubMedGoogle Scholar
- Ramoni MF, Sebastiani P, Kohane IS: Cluster analysis of gene expression dynamics. Proc Natl Acad Sci USA 2002, 99(14):9121–9126. 10.1073/pnas.132656399PubMed CentralView ArticlePubMedGoogle Scholar
- Schliep A, Schonhuth A, Steinhoff C: Using hidden Markov models to analyze gene expression time course data. Bioinformatics 2003, 19(Suppl 1):i255–263. 10.1093/bioinformatics/btg1036View ArticlePubMedGoogle Scholar
- Xu XL, Olson JM, Zhao LP: A regression-based method to identify differentially expressed genes in microarray time course studies and its application in an inducible Huntington's disease transgenic model. Hum Mol Genet 2002, 11(17):1977–1985. 10.1093/hmg/11.17.1977View ArticlePubMedGoogle Scholar
- Fischer EA, Friedman MA, Markey MK: Empirical comparison of tests for differential expression on time-series microarray experiments. Genomics 2007, 89(4):460–470. 10.1016/j.ygeno.2006.10.008View ArticlePubMedGoogle Scholar
- Meissner A, Mikkelsen TS, Gu H, Wernig M, Hanna J, Sivachenko A, Zhang X, Bernstein BE, Nusbaum C, Jaffe DB, Gnirke A, Jaenisch R, Lander ES: Genome-scale DNA methylation maps of pluripotent and differentiated cells. Nature 2008, 454(7205):766–770.PubMed CentralPubMedGoogle Scholar
- Lister R, Pelizzola M, Dowen RH, Hawkins RD, Hon G, Tonti-Filippini J, Nery JR, Lee L, Ye Z, Ngo QM, Edsall L, Antosiewicz-Bourget J, Stewart R, Ruotti V, Millar AH, Thomson JA, Ren B, Ecker JR: Human DNA methylomes at base resolution show widespread epigenomic differences. Nature 2009, 462(7271):315–322. 10.1038/nature08514PubMed CentralView ArticlePubMedGoogle Scholar
- Pleasance ED, Cheetham RK, Stephens PJ, McBride DJ, Humphray SJ, Greenman CD, Varela I, Lin ML, Ordonez GR, Bignell GR, Ye K, Alipaz J, Bauer MJ, Beare D, Butler A, Carter RJ, Chen L, Cox AJ, Edkins S, Kokko-Gonzales PI, Gormley NA, Grocock RJ, Haudenschild CD, Hims MM, James T, Jia M, Kingsbury Z, Leroy C, Marshall J, Menzies A, et al.: A comprehensive catalogue of somatic mutations from a human cancer genome. Nature 2009.Google Scholar
- Shah N, Sukumar S: The Hox genes and their roles in oncogenesis. Nat Rev Cancer 10(5):361–371. 10.1038/nrc2826Google Scholar
- Krivtsov AV, Armstrong SA: MLL translocations, histone modifications and leukaemia stem-cell development. Nat Rev Cancer 2007, 7(11):823–833. 10.1038/nrc2253View ArticlePubMedGoogle Scholar
- Slany RK: The molecular biology of mixed lineage leukemia. Haematologica 2009, 94(7):984–993. 10.3324/haematol.2008.002436PubMed CentralView ArticlePubMedGoogle Scholar
- Li T, Huang J, Jiang Y, Zeng Y, He F, Zhang MQ, Han Z, Zhang X: Multi-stage analysis of gene expression and transcription regulation in C57/B6 mouse liver development. Genomics 2009, 93(3):235–242. 10.1016/j.ygeno.2008.10.006PubMed CentralView ArticlePubMedGoogle Scholar
- Small CL, Shima JE, Uzumcu M, Skinner MK, Griswold MD: Profiling gene expression during the differentiation and development of the murine embryonic gonad. Biol Reprod 2005, 72(2):492–501. 10.1095/biolreprod.104.033696PubMed CentralView ArticlePubMedGoogle Scholar
- Naxerova K, Bult CJ, Peaston A, Fancher K, Knowles BB, Kasif S, Kohane IS: Analysis of gene expression in a developmental context emphasizes distinct biological leitmotifs in human cancers. Genome Biol 2008, 9(7):R108. 10.1186/gb-2008-9-7-r108PubMed CentralView ArticlePubMedGoogle Scholar
- Otu HH, Naxerova K, Ho K, Can H, Nesbitt N, Libermann TA, Karp SJ: Restoration of liver mass after injury requires proliferative and not embryonic transcriptional patterns. J Biol Chem 2007, 282(15):11197–11204. 10.1074/jbc.M608441200View ArticlePubMedGoogle Scholar
- Khetchoumian K, Teletin M, Tisserand J, Mark M, Herquel B, Ignat M, Zucman-Rossi J, Cammas F, Lerouge T, Thibault C, Metzger D, Chambon P, Losson R: Loss of Trim24 (Tif1alpha) gene function confers oncogenic activity to retinoic acid receptor alpha. Nat Genet 2007, 39(12):1500–1506. 10.1038/ng.2007.15View ArticlePubMedGoogle Scholar
- Thomas RS, O'Connell TM, Pluta L, Wolfinger RD, Yang L, Page TJ: A comparison of transcriptomic and metabonomic technologies for identifying biomarkers predictive of two-year rodent cancer bioassays. Toxicol Sci 2007, 96(1):40–46. 10.1093/toxsci/kfl171View ArticlePubMedGoogle Scholar
- Pla P, Hirsch MR, Le Crom S, Reiprich S, Harley VR, Goridis C: Identification of Phox2b-regulated genes by expression profiling of cranial motoneuron precursors. Neural Dev 2008, 3: 14. 10.1186/1749-8104-3-14PubMed CentralView ArticlePubMedGoogle Scholar
- Hendrix ND, Wu R, Kuick R, Schwartz DR, Fearon ER, Cho KR: Fibroblast growth factor 9 has oncogenic activity and is a downstream target of Wnt signaling in ovarian endometrioid adenocarcinomas. Cancer Res 2006, 66(3):1354–1362. 10.1158/0008-5472.CAN-05-3694View ArticlePubMedGoogle Scholar
- Rosen MB, Schmid JE, Das KP, Wood CR, Zehr RD, Lau C: Gene expression profiling in the liver and lung of perfluorooctane sulfonate-exposed mouse fetuses: comparison to changes induced by exposure to perfluorooctanoic acid. Reprod Toxicol 2009, 27(3–4):278–288. 10.1016/j.reprotox.2009.01.007View ArticlePubMedGoogle Scholar
- Thorrez L, Van Deun K, Tranchevent LC, Van Lommel L, Engelen K, Marchal K, Moreau Y, Van Mechelen I, Schuit F: Using ribosomal protein genes as reference: a tale of caution. PLoS One 2008, 3(3):e1854. 10.1371/journal.pone.0001854PubMed CentralView ArticlePubMedGoogle Scholar
- LaVaute TM, Yoo YD, Pankratz MT, Weick JP, Gerstner JR, Zhang SC: Regulation of neural specification from human embryonic stem cells by BMP and FGF. Stem Cells 2009, 27(8):1741–1749. 10.1002/stem.99PubMed CentralView ArticlePubMedGoogle Scholar
- Lu SJ, Hipp JA, Feng Q, Hipp JD, Lanza R, Atala A: GeneChip analysis of human embryonic stem cell differentiation into hemangioblasts: an in silico dissection of mixed phenotypes. Genome Biol 2007, 8(11):R240. 10.1186/gb-2007-8-11-r240PubMed CentralView ArticlePubMedGoogle Scholar
- Pollard SM, Yoshikawa K, Clarke ID, Danovi D, Stricker S, Russell R, Bayani J, Head R, Lee M, Bernstein M, Squire JA, Smith A, Dirks P: Glioma stem cell lines expanded in adherent culture have tumor-specific phenotypes and are suitable for chemical and genetic screens. Cell Stem Cell 2009, 4(6):568–580. 10.1016/j.stem.2009.03.014View ArticlePubMedGoogle Scholar
- Hever A, Roth RB, Hevezi P, Marin ME, Acosta JA, Acosta H, Rojas J, Herrera R, Grigoriadis D, White E, Conlon PJ, Maki RA, Zlotnik A: Human endometriosis is associated with plasma cells and overexpression of B lymphocyte stimulator. Proc Natl Acad Sci USA 2007, 104(30):12451–12456. 10.1073/pnas.0703451104PubMed CentralView ArticlePubMedGoogle Scholar
- Hong Y, Ho KS, Eu KW, Cheah PY: A susceptibility gene set for early onset colorectal cancer that integrates diverse signaling pathways: implication for tumorigenesis. Clin Cancer Res 2007, 13(4):1107–1114. 10.1158/1078-0432.CCR-06-1633View ArticlePubMedGoogle Scholar
- Turashvili G, Bouchal J, Baumforth K, Wei W, Dziechciarkova M, Ehrmann J, Klein J, Fridman E, Skarda J, Srovnal J, Hajduch M, Murray P, Kolar Z: Novel markers for differentiation of lobular and ductal invasive breast carcinomas by laser microdissection and microarray analysis. BMC Cancer 2007, 7: 55. 10.1186/1471-2407-7-55PubMed CentralView ArticlePubMedGoogle Scholar
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.