Systematically characterizing and prioritizing chemosensitivity related gene based on Gene Ontology and protein interaction network
- Xin Chen†1,
- Wei Jiang†1,
- Qianghu Wang1,
- Teng Huang1,
- Peng Wang1,
- Yan Li1,
- Xiaowen Chen1,
- Yingli Lv1 and
- Xia Li1Email author
© Chen et al.; licensee BioMed Central Ltd. 2012
Received: 18 November 2011
Accepted: 27 August 2012
Published: 2 October 2012
The identification of genes that predict in vitro cellular chemosensitivity of cancer cells is of great importance. Chemosensitivity related genes (CRGs) have been widely utilized to guide clinical and cancer chemotherapy decisions. In addition, CRGs potentially share functional characteristics and network features in protein interaction networks (PPIN).
In this study, we proposed a method to identify CRGs based on Gene Ontology (GO) and PPIN. Firstly, we documented 150 pairs of drug-CCRG (curated chemosensitivity related gene) from 492 published papers. Secondly, we characterized CCRGs from the perspective of GO and PPIN. Thirdly, we prioritized CRGs based on CCRGs’ GO and network characteristics. Lastly, we evaluated the performance of the proposed method.
We found that CCRG enriched GO terms were most often related to chemosensitivity and exhibited higher similarity scores compared to randomly selected genes. Moreover, CCRGs played key roles in maintaining the connectivity and controlling the information flow of PPINs. We then prioritized CRGs using CCRG enriched GO terms and CCRG network characteristics in order to obtain a database of predicted drug-CRGs that included 53 CRGs, 32 of which have been reported to affect susceptibility to drugs. Our proposed method identifies a greater number of drug-CCRGs, and drug-CCRGs are much more significantly enriched in predicted drug-CRGs, compared to a method based on the correlation of gene expression and drug activity. The mean area under ROC curve (AUC) for our method is 65.2%, whereas that for the traditional method is 55.2%.
Our method not only identifies CRGs with expression patterns strongly correlated with drug activity, but also identifies CRGs in which expression is weakly correlated with drug activity. This study provides the framework for the identification of signatures that predict in vitro cellular chemosensitivity and offers a valuable database for pharmacogenomics research.
Chemotherapy serves as a general defense against a large number of malignancies. However, only a portion of patients favorably respond to chemotherapy; drug efficacy and adverse drug reactions vary widely among patients [1–3]. Thus it is important to predict chemotherapy response prior to treatment and to select alternative treatment regimens for chemotherapy-resistant patients. A number of potential biomarkers have been identified in previous studies and utilized for patient specific chemotherapy selection . Gene expression profiles of patients pre-treatment have the potential capability to predict responses to chemotherapy; for example, ERCC1 activation is critical in the generation of cisplatin resistance . Asparagine synthetase protein expression measured by immunoassay is a predictor of L-asparaginase activity in ovarian cancer cell lines . Ovarian cancer cell lines that express low ASNS protein levels are generally more sensitive to L-ASP treatment. The expression level of p27 is also a potential candidate predictor for patient selection for rapamycin analogs-based therapy . The National Cancer Institute has used a panel of 60 diverse human cancer cell lines (NCI 60 cell line) (http://genome-www.stanford.edu/nci60/index.shtml) for drug-related research . It was reported that proteomic data solved pharmacologic issues more directly than genomic data . For NCI 60, protein expression levels have been measured for 52 antibodies using reverse-phase protein lysate microarrays . The limited number of proteins restricts identification of chemosensitivity proteins.
Some researchers have devised methods to identify chemosensitivity related genes (CRGs) based on the correlation of gene expression data and drug activity within the NCI 60 dataset [11–14]. Mariadason et al. identified CRGs for 5-fluorouracil (5-FU) by calculating the correlation coefficient of gene expression and 5-FU activity. The 50 most highly correlated genes were used to predict the response to 5-FU . Szakacs et al. coupled gene expression and drug activity with bootstrap analysis to identify gene-drug pairs in which the gene potentially predicts resistance to the drug . Lorenzi et al. reported that correlation coefficient of some drug-gene was not high (r = −0.21). The gene would not be regarded as CRG based on correlation analysis. However, aspargine synthetase was able to predict sensitivity of L-ASP . However, Researchers have developed additional computational methods based on gene expression. Staunton et al. substituted correlation with t-statistics and applied 10-fold cross-validation to define classifiers for each of 232 compounds . Gao et al. identified CRGs by integrating gene expression and transcription factor binding data . Bayesian networks have identified CRGs by integrating different types of data such as gene expression and ChIP-chip data . Although these methods provide vital information regarding CRGs, they consider individual genes in isolation rather than in the context of their functional interactions. In fact, genes are not functionally independent; they work in synergy to perform certain biological functions, such as biological processes, molecular function, complexes or pathways [20–22]. Moreover, it has been reported that chemosensitivity does not appear to be determined by the expression of a single gene . Prediction of CRGs with gene sets is indeed a much more robust method compared to single gene measurement . Taken together, these findings indicate that it is warranted to comprehensively explore biologically significant CRGs by not only considering the correlation between drug activity profiles and gene expression profiles, but by investigating the functional interactions of genes; this could potentially broaden the current understanding of chemosensitivity by elucidation of the context of a functional gene set.
Analyses of protein-protein interaction networks (PPINs) have revealed that genes with high betweenness centrality may be common predictive markers of chemosensitivity . Sensitivity to a variety of compounds may be also influenced by certain aspects of Gene Ontology (GO) functionality, such as cell death, NADH dehydrogenase activity, ABC transporter, cell adhesion, G-protein coupled receptor protein signalling and macromolecule metabolism [16, 24, 26–29]. Previous studies have identified disease genes, radioresistance genes and drug target genes based on Gene Ontology and protein-interaction networks [30–32].
In this study, we proposed a novel method to identify CRGs by integrating information of Gene Ontology, protein interaction network, drug activity profile and gene expression profile. We documented 150 drug-CCRG pairs (curated chemosensitivity related gene) from 492 published papers. Most of the GO terms enriched by CCRGs were related to chemosensitivity and these terms were more similar to each other than random GO terms. Moreover, network analysis indicated that CCRGs exhibited a higher degree and betweenness centrality than random genes. Thus, we constructed an initial drug-candidate CRG network that included two types of nodes: drug nodes, in which activity data were available, and gene nodes in which expression data were available in NCI 60 cell lines. Edges of the network were weighted by Pearson’s correlation coefficient (PCC) between gene expression and drug activity. We then pruned the network using CCRGs’ enriched GO categories and the CCRG network characteristics. Using this method we obtained a database of predicted drug-CRGs.
Curating drug-CCRG pairs
We searched the PubMed database with a list of keywords, such as ‘drug/compound/chemical/small molecule’ and ‘sensitive/sensitivity/resistant/resistance/response’ in the title/abstract, and using ‘National Cancer Institute’ and ‘gene/transcript/protein’ in any field of the literature. The drug-CCRG pairs were derived from experimental studies of NCI 60 cell lines (RT-PCR, siRNA, crystallographic data, etc.); of the 492 retrieved published papers, 150 pairs of drug-CCRG were documented, including 64 drugs and 94 genes. Each entry in the database contained detailed information on a drug-CCRG relationship, including the general name of the drug, gene symbol of CCRG, the cell line where the relationship was documented, literature ID in the NCBI PubMed database, and a brief description of the drug-CCRG relationship. For example, over-expression of Macrophage inhibitory cytokine-1 (MIC-1) predicted sensitivity of ribotoxic anisomycin. The annotated drug-CCRG table is supplemented in Additional file 1.
Drug activity data and gene expression data
The National Cancer Institute's NCI 60 cell line panel is the most extensively characterized set of cells. These 60 human tumor cell lines are derived from patients with leukemia, melanoma, lung, colon, central nervous system, ovarian, renal, breast and prostate cancers. The analysis is presented in terms of drug activity data and microarray-based gene expression profiles of the NCI 60 cell lines.
The drug activity data we utilized included 4463 drugs . Drug activities were recorded across the 60 human cancer cell lines using the logarithm of GI50 to base 10 (lgGI50). GI50 is the concentration required to inhibit cell growth by 50% compared with untreated controls. The activity profile of an agent consists of 60 such activity values, one for each cell line.
NCI 60 cell lines have been subjected to DNA and RNA microarray analysis. We utilized gene expression RNA profile data  (Affy-U133A, GCRMA-normalized), downloaded from the CellMiner database ; it comprises expression patterns of 22283 probes in NCI 60 cell lines.
Correlation of drug activity and gene expression
Among the original 4463 drugs, 19 drugs were discarded because their activity data were missing in more than 80% of the NCI 60 cell lines. Thus the total number of drugs we analyzed in this study was 4444. D represents drug activity profile of the NCI 60 cell lines, each row represents a drug and each column represents a cell line, each element a ij represents the drug activity (GI50) of drug d j in cell line C j , i = 1,2,…,4444, j = 1,2,…,59. G represents the gene expression profiles of the NCI 60 cell lines, each row represents a gene and each column represents a cell line, each element e ij represents the expression level of gene g i in cell line C j , i = 1,2,…,12633. The total number of genes we analyzed in the manuscript was 12633.
where E is expectation, cov is covariance, and X, Y represent a drug and a gene, respectively. δ X 2 = E(X2) − E2(X), δ Y 2 = E(Y2) − E2(Y).
For drug-CCRG pair d2-g1, we calculated the PCC between drug activity of d2 and gene expression of g1 in the NCI 60 cell line. Similarly, we calculated PCC of other drug-CCRG pair. We ranked the absolute PCC of all N drug-CCRG pairs in ascending order and set the PCC threshold as the 5th percentile of N PCCs. Thus, 95% of drug-CCRGs were detected using this threshold.
Constructing the initial drug- candidate CRG network
The initial drug-candidate CRG network includes two types of nodes: drug nodes, all the drugs with available activity data, and gene nodes with available expression data in NCI 60 cell lines. The edges of the network are weighted by Pearson’s correlation coefficient (PCC) between gene expression and drug activity. For some drugs, their activity data are unavailable and represented by NaN. We calculated PCC in the cell lines whose activity data are non-NaN.
GO enrichment using fisher exact test
Illustration of Fisher Exact test
In GO term
Not In GO term
Protein-protein interaction network
A number of publicly available human protein-protein interaction databases have become an important resource for the investigation of biological networks. PPI (protein-protein interaction) data in Human Protein Reference Database (HPRD)  are experimentally derived and manually extracted from the literature by expert biologists who read, interpret and analyze the published data. We downloaded protein interaction data from HPRD on the website http://www.hprd.org/download. The number of binary non-redundant human PPIs is 36687 in HPRD. The number of genes annotated with at least one interaction is 9408. We utilized “MatlabBGL” toolbox (http://dgleich.github.com/matlab-bgl/) and R package “igraph” to calculate network scores .
Characterizing CCRG properties in PPIN
The degree of a gene is the number of its neighborhood genes in PPI network. One gene with high degree, termed a hub gene, plays a key role in maintaining the interactions between this gene and its neighborhood genes.
Betweenness centrality of one gene g is calculated as following:
Where nodes s and t are nodes in the network different from node i in PPI network, d st denotes the number of shortest paths from s to t, δst(i) is the number of shortest path from s to t that i lies on. For two genes s and t, the ratio is the number of shortest path that g lies on relative to all the possible shortest paths between genes s and t. The sum of the ratio of all gene pairs is betweenness centrality of gene g. If one gene exhibits high betweenness centrality, it is likely to play a vital role in gene communication and is termed a bottleneck gene.
Qstatistics to integrate ranks from multiple data resources
The receiver operating characteristic (ROC) curve was used to assess the performance of the two methods: the proposed method that integrates gene expression and functional interaction, and the other method based on gene expression. We ranked all CRGs in both methods and determined whether CCRGs ranked at the top of the list. Each gene was ranked in the order of degree and betweenness centrality, respectively. Next, we utilized Q statistic to integrate the two ranks into a final rank. The details are described as follows: , where r i is the rank ratio for data source i, N is the number of data sources used, and r 0 = 0. In the proposed method, N = 2.
Correlation-based analysis of the drug-CCRG pairs
Figure 2 and Figure 3 show that the majority of drug-CCRGs exhibit a low correlation between gene expression and drug activity. Moreover, 27/62 (44%) of drug-CCRG correlations tend to be random by comparing z i with z threshold . Thus we investigated to integrate additional functional information to predict drug-CRGs.
GO enrichment analysis of CCRGs
CCRGs are significantly enriched in 204 terms (p < 0.01) according to Fisher’s exact test. For a complete list of enriched GO terms, see Additional file 3. The majority of enriched GO terms are related to chemosensitivity. For example, the GO terms “basolateral plasma membrane” are related to chemosensitivity linked by ABCB5 . First-pass elimination of CRC 220 is due to an active carrier-mediated transport process in the “basolateral plasma membrane” . Lesions in oncogenes and tumour suppressor genes involved in “the regulation of programmed cell death” appear to be important in the evolution of drug resistance . Proteins involved in “regulation of apoptosis” are associated with cisplatin chemosensitivity in germ cell tumors . Genes involved in “regulation of cell cycle”, such as p53 protein family, contribute to chemotherapeutic drug response in gastrointestinal tumors . “Xenobiotic metabolism” involves modifying the chemical structure of xenobiotics, such as drugs and poisons. Reactions in these pathways contribute to chemosensitivity in cancer. Furthermore, CCRG enriched GO terms exhibit significantly greater similarity compared to randomly selected genes. This indicates that CCRG enriched GO terms are more similar to each other when compared with GO terms where random genes enriched (Additional file 4).
The characteristics of CCRGs in PPIN
Degree of CCRG compared with random genes
mean of CCRG
mean of random genes
Betweenness centrality of CCRG compared with random genes
mean of CCRG
mean of random gene
Performance of the proposed method to identify drug-CRGs
Performance of our method to predict drug-CRGs under different thresholds
Threshold of degree*
Threshold of betweenness centrality
The proposed method
Method based on gene expression
Number of identified CCRGs
Number of identified CCRGs#
We next evaluated the performance of the proposed method by ROC to determine whether CCRGs were distinguished from other genes. For the proposed method, we ranked all of the genes in predicted drug-CRGs using the Q statistic (See details in Methods) in order to integrate various separate data sources. We integrated ranks of degree and betweenness centrality to determine whether CCRGs ranked at the top of the list. According to Q statistics and whether genes were CCRGs, we plotted the ROC curves. For traditional correlation method, we ranked all drug-CRG pairs using absolute PCC of gene expression and drug activity. According to PCC and whether genes were CCRGs, we also plotted the ROC curves.
Identification of CRGs by integrating CCRGs’ properties in GO and PPIN
Based on gene expression, GO categories, and network characteristics, we identified CRGs for drugs. Combined filtering method is superior compared with the method using only Pearson’s correlation coefficients based on gene expression. We used this combined filtering method to identify CRGs for all of the drugs, whose activities were screened in NCI 60 cell lines. Consequently, we obtained 53 genes that were not only associated with chemosensitivity related GO categories but also played key roles in maintaining connectivity and controlling the information flow of PPIN. Among the 53 CRGs, 32 were previously reported as chemosensitivity related genes. The full gene list is in Additional file 6.
Our findings are supported by previous studies. Genes with high correlation coefficients are identified as CRGs. For example, EGFR is negatively correlated with activity of Tamoxifen, and the Pearson’s correlation coefficient (PCC) is – 0.39. This suggests that expression of EGFR can predict the resistance to Tamoxifen, which is consistent with a previous study in which EGFR product resulted in decreased susceptibility to Tamoxifen . At the same time, BRCA1 is positively correlated with activity of Tamoxifen (PCC = 0.25); this indicates that BRCA1 expression can predict sensitivity of Tamoxifen, which is in concordance with a previous study in which the overexpression of BRCA1 results in increased susceptibility to Tamoxifen. We also identified candidate CRGs with low PCC. For example, although AKT1 is weakly correlated with sensitivity of Doxorubicin (PCC = 0.13), it has been reported to result in increased susceptibility to Doxorubicin . EGFR product affects the susceptibility to Fluorouracil (PCC=– 0.2) , RB1 affects the susceptibility to Fluorouracil (PCC=– 0.09) , RELA product affects the susceptibility to Doxorubicin (PCC=– 0.05) , STAT3 affects the susceptibility to Fluorouracil (PCC=– 0.18) , and TP53 product affects the susceptibility to Fluorouracil (PCC = 0.04) . These results indicate that these genes exhibit the potential to predict chemosensitivity of drugs before initiating therapy, which could potentially aid clinical decisions and allow for more individualized treatment strategies for patients.
The high-resolution profiling at the mRNA level and high-throughput drug sensitivity data of NCI 60 allow for comprehensively mapping of mRNA profiles for molecular pharmacologic and drug discovery . There are previously reported high-throughput studies on CRG identification for drugs; however, most of these studies are based on gene expression. Some studies reported genes with expression levels highly correlated with drug activity as CRGs, chemosensitivity genes with low PCC were excluded. Aside from correlation analysis, some researchers have developed other computational methods based on gene expression. However, individual genes were studied in isolation rather than in the context of their functional interactions. In fact, genes are not functionally independent; they work in synergy to perform biological function.
In our proposed method, we utilized high-throughput gene expression profiles to predict CRGs by integrating drug-gene correlations, gene function annotation, and network information. We systematically characterized CCRGs in the context of functional genomic data; we then prioritized CRGs based on these CCRG characteristics. Firstly, we conducted an extensive literature survey and manually curated a compendium of CCRGs. According to GO analysis on three ontologies, most of the CCRG enriched GO terms were related to chemosensitivity. Moreover, these GO terms were more similar to each other compared to randomly selected genes. CCRGs also play key roles in protein-protein interaction network (PPIN). They control the information flow of PPIN and maintain connectivity of PPIN. The initial drug-candidate CRG network was pruned according to these characteristics; consequently we obtained a database of predicted drug-CRGs for all drugs whose activity profiles were screened in NCI 60 cell lines. The results demonstrated that our method can not only identify CRGs whose expression is strongly correlated with drug activity, but also can identify CRGs whose expression is weakly correlated with drug activity. These results are powerfully supported by previous studies. From the predicted drug-CRGs, the researchers can easily access genes and drugs of interest, thus facilitating further studies. Functional genomic information, such as GO categories and protein interaction networks, aid the identification of CRGs unable to be identified by methods based only on similarity between gene expressions and drug activity.
The present analysis has the following limitations: (a) the drug-CCRGs we curated are limited to NCI 60 data. (b) the data presented here give an incomplete biological picture of the relationship between drug and CRG. Further validation of drug-CRG relationships is necessary prior to clinical application. (c) the conclusions were extrapolated from in vitro to in vivo. Transformed cell lines might further evolve in vitro and might not reflect the tumor from which they were originally isolated. (d) finally, the relationships established between drug activities and gene expression levels are correlative, not causal.
In summary, we provide an integrated method of identifying CRGs that combines gene expression, drug activity data and functional information for genes such as GO categories and PPIN. We documented 150 pairs of drug-CCRG from 492 published papers. CCRG enriched GO terms were generally related to chemosensitivity. These GO terms exhibited higher similarity compared to GO terms enriched by randomly selected genes. Moreover, CCRGs play key roles in maintaining connectivity and controlling information flow of PPIN. Thus, we pruned the initial drug-candidate CRG network based on CCRG GO categories and network characteristics. As a result, we obtained a database of predicted drug-CRGs. It includes 53 CRGs, 32 of which have been previously reported to be chemosensitivity related genes.
The CRGs identified will potentially allow for greater treatment efficacy and fewer unnecessary side effects. For patients predicted not to respond to certain agent, alternative agents or combined agents could be considered. Candidate second-line anticancer drugs for combination therapy may be selected based on the database of predicted drug-CRGs. Moreover, the CRGs may serve as candidate drug targets for the development of new drugs. With additional validated drug-CCRG pairs, our proposed method could potentially provide valuable resources for pharmacogenomics research and contribute to the framework for individualized medicine.
This work is supported in part by the National Natural Science Foundation of China (Grant Nos. 30900837, 31100948, 61073136, 30871394,), the National High Tech Development Project of China, the 863 Program (Grant Nos. 2007AA02Z329) and the National Science Foundation of Heilongjiang Province (Grant Nos. ZD200816-01, ZJG0501).
- Cheok MH, Evans WE: Acute lymphoblastic leukaemia: a model for the pharmacogenomics of cancer therapy. Nat Rev Cancer. 2006, 6: 117-129.View ArticlePubMed
- Fellenberg J, Dechant MJ, Ewerbeck V, Mau H: Identification of drug-regulated genes in osteosarcoma cells. Int J Cancer. 2003, 105: 636-643.View ArticlePubMed
- Huang Y: Pharmacogenetics/genomics of membrane transporters in cancer chemotherapy. Cancer Metastasis Rev. 2007, 26: 183-201.View ArticlePubMed
- Shepherd FA, Rosell R: Weighing tumor biology in treatment decisions for patients with non-small cell lung cancer. J Thorac Oncol. 2007, 2 (Suppl 2): S68-S76.View ArticlePubMed
- Hsu DS, Lan HY, Huang CH, Tai SK, Chang SY, Tsai TL, Chang CC, Tzeng CH, Wu KJ, Kao JY, Yang MH: Regulation of excision repair cross-complementation group 1 by Snail contributes to cisplatin resistance in head and neck cancer. Clin Cancer Res. 2010, 16: 4561-4571.View ArticlePubMed
- Lorenzi PL, Llamas J, Gunsior M, Ozbun L, Reinhold WC, Varma S, Ji H, Kim H, Hutchinson AA, Kohn EC, Goldsmith PK, Birrer MJ, Weinstein JN: Asparagine synthetase is a predictive biomarker of L-asparaginase activity in ovarian cancer cell lines. Mol Cancer Ther. 2008, 7: 3123-3128.PubMed CentralView ArticlePubMed
- Chen G, Yang N, Wang X, Zheng SY, Chen Y, Tong LJ, Li YX, Meng LH, Ding J: Identification of p27/KIP1 expression level as a candidate biomarker of response to rapalogs therapy in human cancer. J Mol Med (Berl). 2010, 88: 941-952.View Article
- Shoemaker RH: The NCI60 human tumour cell line anticancer drug screen. Nat Rev Cancer. 2006, 6: 813-823.View ArticlePubMed
- Ma Y, Ding Z, Qian Y, Shi X, Castranova V, Harner EJ, Guo L: Predicting cancer drug response by proteomic profiling. Clin Cancer Res. 2006, 12: 4583-4589.View ArticlePubMed
- Nishizuka S, Charboneau L, Young L, Major S, Reinhold WC, Waltham M, Kouros-Mehr H, Bussey KJ, Lee JK, Espina V, Munson PJ, Petricoin E, Liotta LA, Weinstein JN: Proteomic profiling of the NCI 60 cancer cell lines using new high-density reverse-phase lysate microarrays. Proc Natl Acad Sci USA. 2003, 100: 14229-14234.PubMed CentralView ArticlePubMed
- Paull KD, Shoemaker RH, Hodes L, Monks A, Scudiero DA, Rubinstein L, Plowman J, Boyd MR: Display and analysis of patterns of differential activity of drugs against human tumor cell lines: development of mean graph and COMPARE algorithm. J Natl Cancer Inst. 1989, 81: 1088-1092.View ArticlePubMed
- Shankavaram UT, Reinhold WC, Nishizuka S, Major S, Morita D, Chary KK, Reimers MA, Scherf U, Kahn A, Dolginow D, Cossman J, Kaldjian EP, Scudiero DA, Petricoin E, Liotta L, Lee JK, Weinstein JN: Transcript and protein expression profiles of the NCI 60 cancer cell panel: an integromic microarray study. Mol Cancer Ther. 2007, 6: 820-832.View ArticlePubMed
- Bussey KJ, Chin K, Lababidi S, Reimers M, Reinhold WC, Kuo WL, Gwadry F, Kouros-Mehr H, Fridlyand J, Jain A, Collins C, Nishizuka S, Tonon G, Roschke A, Gehlhaus K, Kirsch I, Scudiero DA, Gray JW, Weinstein JN, Ajay: Integrating data on DNA copy number with gene expression levels and drug sensitivities in the NCI 60 cell line panel. Mol Cancer Ther. 2006, 5: 853-867.PubMed CentralView ArticlePubMed
- Scherf U, Ross DT, Waltham M, Smith LH, Lee JK, Tanabe L, Kohn KW, Reinhold WC, Myers TG, Andrews DT, Scudiero DA, Eisen MB, Sausville EA, Pommier Y, Botstein D, Brown PO, Weinstein JN: A gene expression database for the molecular pharmacology of cancer. Nat Genet. 2000, 24: 236-244.View ArticlePubMed
- Mariadason JM, Arango D, Shi Q, Wilson AJ, Corner GA, Nicholas C, Aranes MJ, Lesser M, Schwartz EL, Augenlicht LH: Gene expression profiling-based prediction of response of colon carcinoma cells to 5-fluorouracil and camptothecin. Cancer Res. 2003, 63: 8791-8812.PubMed
- Szakacs G, Annereau JP, Lababidi S, Shankavaram U, Arciello A, Bussey KJ, Reinhold W, Guo Y, Kruh GD, Reimers M, Weinstein JN, Gottesman MM: Predicting drug sensitivity and resistance: profiling ABC transporter genes in cancer cells. Cancer Cell. 2004, 6: 129-137.View ArticlePubMed
- Staunton JE, Slonim DK, Coller HA, Tamayo P, Angelo MJ, Park J, Scherf U, Lee JK, Reinhold WO, Weinstein JN, Mesirov JP, Lander ES, Golub TR: Chemosensitivity prediction by transcriptional profiling. Proc Natl Acad Sci USA. 2001, 98: 10787-10792.PubMed CentralView ArticlePubMed
- Gao F, Foat BC, Bussemaker HJ: Defining transcriptional networks through integrative modeling of mRNA expression and transcription factor binding data. BMC Bioinforma. 2004, 5: 31.View Article
- Liu X, Jessen WJ, Sivaganesan S, Aronow BJ, Medvedovic M: Bayesian hierarchical model for transcriptional module discovery by jointly modeling gene expression and ChIP-chip data. BMC Bioinforma. 2007, 8: 283.View Article
- Kanehisa M, Goto S: KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 2000, 28: 27-30.PubMed CentralView ArticlePubMed
- Ruepp A, Brauner B, Dunger-Kaltenbach I, Frishman G, Montrone C, Stransky M, Waegele B, Schmidt T, Doudieu ON, Stumpflen V, Mewes HW: CORUM: the comprehensive resource of mammalian protein complexes. Nucleic Acids Res. 2008, 36: D646-D650.PubMed CentralView ArticlePubMed
- Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM, Davis AP, Dolinski K, Dwight SS, Eppig JT, Harris MA, Hill DP, Issel-Tarver L, Kasarskis A, Lewis S, Matese JC, Richardson JE, Ringwald M, Rubin GM, Sherlock G: Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat Genet. 2000, 25: 25-29.PubMed
- Dan S, Tsunoda T, Kitahara O, Yanagawa R, Zembutsu H, Katagiri T, Yamazaki K, Nakamura Y, Yamori T: An integrated database of chemosensitivity to 55 anticancer drugs and gene expression profiles of 39 human cancer cell lines. Cancer Res. 2002, 62: 1139-1147.PubMed
- Kutalik Z, Beckmann JS, Bergmann S: A modular approach for integrative analysis of large-scale gene-expression and drug-response data. Nat Biotechnol. 2008, 26: 531-539.View ArticlePubMed
- Ragusa M, Avola G, Angelica R, Barbagallo D, Guglielmino MR, Duro LR, Majorana A, Statello L, Salito L, Consoli C, Camuglia MG, Di Pietro C, Milone G, Purrello M: Expression profile and specific network features of the apoptotic machinery explain relapse of acute myeloid leukemia after chemotherapy. BMC Cancer. 2010, 10: 377.PubMed CentralView ArticlePubMed
- Thewes V, Orso F, Jager R, Eckert D, Schafer S, Kirfel G, Garbe S, Taverna D, Schorle H: Interference with activator protein-2 transcription factors leads to induction of apoptosis and an increase in chemo- and radiation-sensitivity in breast cancer cells. BMC Cancer. 2010, 10: 192.PubMed CentralView ArticlePubMed
- Di Camillo B, Sanavia T, Iori E, Bronte V, Roncaglia E, Maran A, Avogaro A, Toffolo G, Cobelli C: The transcriptional response in human umbilical vein endothelial cells exposed to insulin: a dynamic gene expression approach. PLoS One. 2010, 5: e14390.PubMed CentralView ArticlePubMed
- Zhou X, Su Z, Sammons RD, Peng Y, Tranel PJ, Stewart CN, Yuan JS: Novel software package for cross-platform transcriptome analysis (CPTRA). BMC Bioinforma. 2009, 10 (Suppl 11): S16.View Article
- Westhoff MA, Zhou S, Bachem MG, Debatin KM, Fulda S: Identification of a novel switch in the dominant forms of cell adhesion-mediated drug resistance in glioblastoma cells. Oncogene. 2008, 27: 5169-5181.View ArticlePubMed
- Ortutay C, Vihinen M: Identification of candidate disease genes by integrating Gene Ontologies and protein-interaction networks: case study of primary immunodeficiencies. Nucleic Acids Res. 2009, 37: 622-628.PubMed CentralView ArticlePubMed
- Guo Y, Zhu XD, Qu S, Li L, Su F, Li Y, Huang ST, Li DR: Identification of genes involved in radioresistance of nasopharyngeal carcinoma by integrating gene ontology and protein-protein interaction networks. Int J Oncol. 2012, 40: 85-92.PubMed
- Kaimal V, Sardana D, Bardes EE, Gudivada RC, Chen J, Jegga AG: Integrative systems biology approaches to identify and prioritize disease and drug candidate genes. Methods Mol Biol. 2011, 700: 241-259.View ArticlePubMed
- Blower PE, Yang C, Fligner MA, Verducci JS, Yu L, Richman S, Weinstein JN: Pharmacogenomic analysis: correlating molecular substructure classes with microarray gene expression data. Pharmacogenomics J. 2002, 2: 259-271.View ArticlePubMed
- Shankavaram UT, Varma S, Kane D, Sunshine M, Chary KK, Reinhold WC, Pommier Y, Weinstein JN: Cell Miner: a relational database and query tool for the NCI 60 cancer cell lines. BMC Genomics. 2009, 10: 277.PubMed CentralView ArticlePubMed
- Lv S, Li Y, Wang Q, Ning S, Huang T, Wang P, Sun J, Zheng Y, Liu W, Ai J, Li X: A novel method to quantify gene set functional association based on gene ontology. J R Soc Interface. 2011, 9: 1063-1072.PubMed CentralView ArticlePubMed
- da Huang W, Sherman BT, Lempicki RA: Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protoc. 2009, 4: 44-57.View ArticlePubMed
- da Huang W, Sherman BT, Lempicki RA: Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists. Nucleic Acids Res. 2009, 37: 1-13.View ArticlePubMed
- Keshava Prasad TS, Goel R, Kandasamy K, Keerthikumar S, Kumar S, Mathivanan S, Telikicherla D, Raju R, Shafreen B, Venugopal A, Balakrishnan L, Marimuthu A, Banerjee S, Somanathan DS, Sebastian A, Rani S, Ray S, Harrys Kishore CJ, Kanth S, Ahmed M, Kashyap MK, Mohmood R, Ramachandra YL, Krishna V, Rahiman BA, Mohan S, Ranganathan P, Ramabadran S, Chaerkady R, Pandey A: Human Protein Reference Database--2009 update. Nucleic Acids Res. 2009, 37: D767-D772.PubMed CentralView ArticlePubMed
- Csárdi G, Nepusz T: The igraph software package for complex network research. InterJournal. 2006, -Complex Systems: 1695
- Lionetti M, Biasiolo M, Agnelli L, Todoerti K, Mosca L, Fabris S, Sales G, Deliliers GL, Bicciato S, Lombardi L, Bortoluzzi S, Neri A: Identification of microRNA expression patterns and definition of a microRNA/mRNA regulatory network in distinct molecular groups of multiple myeloma. Blood. 2009, 114: e20-e26.View ArticlePubMed
- Choi JK, Yu U, Yoo OJ, Kim S: Differential coexpression analysis using microarray data and its application to human cancer. Bioinformatics. 2005, 21: 4348-4355.View ArticlePubMed
- Wielinga PR, de Waal E, Westerhoff HV, Lankelma J: In vitro transepithelial drug transport by on-line measurement: cellular control of paracellular and transcellular transport. J Pharm Sci. 1999, 88: 1340-1347.View ArticlePubMed
- Eckhardt U, Stüber W, Dickneite G, Reersf M, Petzinger E: First-pass elimination of a peptidomimetic thrombin inhibitor is due to carrier-mediated uptake by the liver. Interaction with bile acid transport systems. Biochem Pharmacol. 1996, 52: 85-96.View ArticlePubMed
- Harrington EA, Fanidi A, Evan GI: Oncogenes and cell death. Curr Opin Genet Dev. 1994, 4: 120-129.View ArticlePubMed
- Mayer F, Stoop H, Scheffer GL, Scheper R, Oosterhuis JW, Looijenga LH, Bokemeyer C: Molecular determinants of treatment response in human germ cell tumors. Clin Cancer Res. 2003, 9: 767-773.PubMed
- Vilgelm AE, Washington MK, Wei J, Chen H, Prassolov VS, Zaika AI: Interactions of the p53 protein family in cellular stress response in gastrointestinal tumors. Mol Cancer Ther. 2010, 9: 693-705.PubMed CentralView ArticlePubMed
- Chu I, Blackwell K, Chen S, Slingerland J: The dual ErbB1/ErbB2 inhibitor, lapatinib (GW572016), cooperates with tamoxifen to inhibit both cell proliferation- and estrogen-dependent gene expression in antiestrogen-resistant breast cancer. Cancer Res. 2005, 65: 18-25.PubMed
- Wen J, Li R, Lu Y, Shupnik MA: Decreased BRCA1 confers tamoxifen resistance in breast cancer cells by altering estrogen receptor-coregulator interactions. Oncogene. 2009, 28: 575-586.PubMed CentralView ArticlePubMed
- Li S, Zhou Y, Wang R, Zhang H, Dong Y, Ip C: Selenium sensitizes MCF-7 breast cancer cells to doxorubicin-induced apoptosis through modulation of phospho-Akt and its downstream substrates. Mol Cancer Ther. 2007, 6: 1031-1038.View ArticlePubMed
- Wu X, Deng Y, Wang G, Tao K: Combining siRNAs at two different sites in the EGFR to suppress its expression, induce apoptosis, and enhance 5-fluorouracil sensitivity of colon cancer cells. J Surg Res. 2007, 138: 56-63.View ArticlePubMed
- Mayhew CN, Perkin LM, Zhang X, Sage J, Jacks T, Knudsen ES: Discrete signaling pathways participate in RB-dependent responses to chemotherapeutic agents. Oncogene. 2004, 23: 4107-4120.View ArticlePubMed
- Tapia MA, Gonzalez-Navarrete I, Dalmases A, Bosch M, Rodriguez-Fanjul V, Rolfe M, Ross JS, Mezquita J, Mezquita C, Bachs O, Gascon P, Rojo F, Perona R, Rovira A, Albanell J: Inhibition of the canonical IKK/NF kappa B pathway sensitizes human cancer cells to doxorubicin. Cell Cycle. 2007, 6: 2284-2292.View ArticlePubMed
- Yu LF, Cheng Y, Qiao MM, Zhang YP, Wu YL: Activation of STAT3 signaling in human stomach adenocarcinoma drug-resistant cell line and its relationship with expression of vascular endothelial growth factor. World J Gastroenterol. 2005, 11: 875-879.PubMed CentralView ArticlePubMed
- Ganzinelli M, Carrassa L, Crippa F, Tavecchio M, Broggini M, Damia G: Checkpoint kinase 1 down-regulation by an inducible small interfering RNA expression system sensitized in vivo tumors to treatment with 5-fluorouracil. Clin Cancer Res. 2008, 14: 5131-5141.View ArticlePubMed
- Weinstein JN, Myers TG, O'Connor PM, Friend SH, Fornace AJ, Kohn KW, Fojo T, Bates SE, Rubinstein LV, Anderson NL, Buolamwini JK, van Osdol WW, Monks AP, Scudiero DA, Sausville EA, Zaharevitz DW, Bunow B, Viswanadhan VN, Johnson GS, Wittes RE, Paull KD: An information-intensive approach to the molecular pharmacology of cancer. Science. 1997, 275: 343-349.View ArticlePubMed
- The pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/1755-8794/5/43/prepub
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.