Cross talk of chromosome instability, CpG island methylator phenotype and mismatch repair in colorectal cancer
- Authors:
- Published online on: May 31, 2018 https://doi.org/10.3892/ol.2018.8860
- Pages: 1736-1746
-
Copyright: © Zhang et al. This is an open access article distributed under the terms of Creative Commons Attribution License.
Abstract
Introduction
Colorectal cancer is one of the most common cancer with leading cause of death (1). Its classical molecular events have been well-studied. The oncogenes in colorectal cancer are ras, scr and c-myc while the tumor suppressor genes are APC and p53. The Wnt pathway is considered to be important in the tumorgenesis of colorectal cancer. In 1990, Fearon and Vogelstein (2) proposed a famous model of colorectal cancer which believes a serials of gene and signaling pathway alterations contribute to the histology changes from normal tissue to adenoma and then to carcinoma. Li et al found that at each stage of colorectal cancer, their gene expression profiles were different (3). Jiang et al found that the early stage colorectal cancer biomarkers and late stage biomarkers were different and they can be connected by signal propagation on the network (4). Many genes were found to be associated with colorectal cancer by gene expression and network analysis (5,6). And many signaling pathways, such as Wnt/β-catenin signaling, epidermal growth factor receptor/Ras signaling, p53 signaling, Notch signaling, Hedgehog signaling, and Hippo signaling, were found to play roles in colorectal cancer (7).
To summary the current understandings of colorectal cancer, there are major mechanisms for colorectal cancer: (1) chromosome instability (CIN), (2) CpG island methylator phenotype (CIMP) and (3) mismatch repair (MMR). In approximately 85% of colorectal cancer patients, the chromosomal instability (CIN) is observed (8). They exhibited genomic instability on the chromosomal level. The CIN patients usually have the poorest prognosis (9). In approximately 15–20% colorectal cancer patients, there are widespread CIMP (10). In approximately 15% colorectal cancer patients, Microsatellite instability (MSI) is detected (11). It is caused by the loss of DNA MMR activity. The MSI patients tend to have a good prognosis (12). These mechanisms are not mutually exclusive. For example, the MMR patients usually also show varying degrees of CIN (8). Different pathways that were used for characterizing each mechanism actually can interact with each other and cross talk (7). Multiple signaling pathways share transcription factors, microRNAs and ligases, such as miR-21, miR-145, FBXW7 and β-TrCP (7).
To systematically investigate the relationship between CIN, CIMP and MMR, we analyzed the gene expression profiles of 585 colorectal cancer patients. These patients were annotated with CIN, CIMP and MMR status. For each status, we applied advanced minimal redundancy maximal relevance (mRMR) and incremental feature selection (IFS) method to select its biomarkers genes. Then we overlapped the CIN, CIMP and MMR biomarker genes. Since they may not directly interact with each other, we used random walk with restart (RWR) method to find the region that the CIN, CIMP and MMR biomarker genes affect and investigated the commonly regulated genes by CIN, CIMP and MMR. The biological functions of these commonly regulated genes were analyzed. Our work found the molecular cross talk among CIN, CIMP and MMR, revealed the internal logic of colorectal tumorgenesis, and provided the emerging therapeutic targets that may be suitable for most colorectal cancer patients rather than a small proportion of patients.
Materials and methods
The gene expression profiles of 585 colorectal cancer patients
We downloaded the gene expression profiles of 585 colorectal cancer patients from GEO (Gene Expression Omnibus) with accession number of GSE39582 (13). The expression levels were measured with Affymetrix Human Genome U133 Plus 2.0 Array which had 54,675 probes corresponding to 20,502 genes. The probes corresponding to the same gene were averaged. The gene expression data was preprocessed with quantile normalization. Within the 585 colon patients, there were 369 CIN+ and 112 CIN-, 93 CIMP+ and 420 CIMP-, 77 dMMR and 459 pMMR. For each analysis, the patients with missing status were excluded. For example, for CIN+ and CIN-comparison, the 369 CIN+ and 112 CIN-patients were considered while 104 without CIN information were excluded.
The CIN-associated gene selection
mRMR gene ranking
We used the mRMR method (14) to rank the genes based on their relevance with CIN status and their redundancy between genes. The mRMR method is based on information theory and has been widely used in bioinformatics filed (15–19). To apply mRMR method, we used the C/C++ version mRMR software downloaded from http://home.penglab.com/proj/mRMR/. With mRMR method, we obtained a ranked gene list. The top 500 mRMR genes were analyzed.
IFS
To determine how many genes should be selected from the mRMR gene list, we adopted the IFS method (4,20–24) and constructed 500 support vector machine (SVM) classifiers. In this study, we used the svm function with default parameters from R package e10171 (https://cran.r-project.org/web/packages/e1071/) to build the SVM classifier. Each time, the top k genes in the mRMR list was used to build the SVM classifier. And the performance of the top k-gene classifier was evaluated with leave-one-out cross validation (LOOCV). To objectively evaluate the classifier's performance, Sensitivity (Sn), Specificity (Sp), Accuracy (ACC) and Mathew's correlation coefficient (MCC) were calculated:
Sn=TPTP+FN Sp=TNTN+FP ACC=TP+TNTP+TN+FP+FN MCC=TP×TN-FP×FN(TP+FP)(TP+FN)(TN+FP)(TN+FN)where TP, TN, FP and FN stand for true positive (CIN+), true negative (CIN-), false positive (CIN+) and false negative (CIN-), respectively. Since the sizes of positive (CIN+) and negative (CIN-) samples were imbalance in this study, MCC which considered both Sn and Sp, was choose as the major measurement (25). At last, based on the IFS curve in which the number of top genes that were used as x-axis and the LOOCV MCCs of classifiers as y-axis, we can decide how many genes should be used to build a classifier with great performance and small complexity. The peak or the change point of the IFS curve were usually chosen.
The CIMP-associated gene selection
Similarly, we can identify the CIMP-associated genes using mRMR and IFS methods. Since the sample size of CIMP+ and CIMP-patients were also imbalance, the MCC was considered as the key measurement for prediction performance evaluation and was used to plot the IFS curve.
The MMR-associated gene selection
Similarly, we can identify the MMR-associated genes by analyzing the gene expression profiles pMMR and dMMR patients using mRMR and IFS methods. The dMMR and pMMR were considered as positive and negative samples, respectively. The MCC was used to plot the IFS curve since there were much more pMMR than dMMR.
The overlapped genes and common downstream genes of CIN, CIMP and MMR
We would like to known whether there is a general mechanism for CIN, CIMP and MMR. The direct way is to overlap the mRMR and IFS identified CIN associated genes, CIMP associated genes and MMR associated genes.
Since the identified CIN associated genes, CIMP associated genes and MMR associated genes may be incomplete or locate at the upstream of the colorectal cancer signaling pathway, we tried to pin down the area affected by the CIN associated genes, CIMP associated genes and MMR associated genes on the protein-protein interaction network of using RWR method (26–29). The STRING network (version 10.0) (30) is a comprehensive protein-protein functional association network that has been widely used (26,28,31–39). It included 19,247 proteins and 4,274,001 interactions. We constructed the network using the protein-protein interactions with confidence score >0.900 which is the highest confidence interaction in STRING database. Then the n*n adjacent matrix (A) of the network which included n proteins was column-wise normalized to make the column sum to be 1 by assign 1/m to the m interaction proteins of protein j in column j and 0 to other proteins without interactions.
The random walk procedure repeat in every time tick (t→t+1) from the initial seed genes which were represented as a n length vector with P0 value of 1/k for the k seed genes and value of 0 for other n-k non-seed genes. The state probabilities Pt+1 at time t+1 is calculated as follow: Pt+1=(1-r)APt+rP0 (5), where Pt is state probabilities at time t, r is the restart probability which is set to 0.7 as suggested by previous studies (26–29,40). It has been reported that if r is in a sizable range (0.5–0.8), the results will have little difference (40). These random walk process will stop when the difference between two steps is smaller than 1e-6. At last, all genes on the network will be assigned with a RWR score which corresponds to the probability of being expanded from the seed genes.
To statically evaluate the significance of RWR score, we randomly chosen the same number of seed genes and calculated their RWR scores for 1,000 times. The significance of actual RWR score can be defined as a permutation P-value of how times the random RWR scores was greater than the actual RWR score over the permutation times which was 1,000 in this study. The genes with permutation P-value smaller than 0.05 were considered as significant RWR expanded genes.
The RWR expanded genes can represent the downstream genes of CIN, CIMP and MMR and be used for common downstream gene analysis. The functions of the common CIN, CIMP and MMR downstream genes were enriched onto KEGG pathways and Gene Ontology (GO) terms using hypergeometric test.
Results and Discussion
The CIN associated genes identified with mRMR and IFS
The top 500 most discriminative genes between CIN+ and CIN-samples were ranked using the mRMR method which considered both their relevance with CIN status, and their redundancy with selected genes. After the genes were ranked by mRMR, we chosen the number of top genes by applying the IFS procedure. Different number of top genes were tried and their prediction performance were evaluated. The IFS curve with the number of genes as x-axis and leave one out cross validation MCC as y-axis was shown in Fig. 1A. It can be seen that when 34 genes were used, the leave one out cross validation MCC was the highest. The leave one out cross validation Sn, Sp, ACC and MCC of these 34 genes were 0.932, 0.696, 0.877 and 0.648, respectively. Therefore these 34 genes were chosen and shown in Table I. As shown in Fig. 2A, the 34 CIN associated genes can cluster the CIN+ and CIN-patients into the right groups. IVD, NDUFAF1, OIP5-AS1, EXOSC9, HSPA4L, RPL22L1, EMC6, NCBP3, CYB5D1, PRPSAP2, RALBP1, ATP9B, ADGRG6, TRIM7, NLRX1, RNF145, CTC1, TMEM102 were highly expressed in CIN-patients while TGFBR2, HERPUD2, KBTBD2, ROCK2, TUFT1, TMEM176A, RHEB, SERINC3, STX16, COMMD7, DYNLRB1, RTFDC1, EIF6, TM9SF4, HEATR4, RRNAD1 were highly expressed in CIN+ patients.
The CIMP associated genes identified with mRMR and IFS
Similarly, the CIMP associated genes can be identified using mRMR and IFS methods. As a result, 19 genes were selected based on the IFS curve shown in Fig. 1B and listed in Table II. The 19 genes' leave one out cross validation Sn, Sp, ACC and MCC were 0.710, 0.976, 0.928 and 0.744, respectively. As shown in Fig. 2B, the 19 CIMP associated genes can cluster the CIMP+ and CIMP-patients into the right groups. VANGL2, ZNF665, JUN, FAM84A, ZBTB38, GRM8, DUSP18, PRDX5, HUNK, QPRT, ZNF141, MLH1, MTERF1 were highly expressed in CIMP-patients while PIWIL1, ADGRG6, FOXD1, HOXC6, AFAP1-AS1, HS3ST1 were highly expressed in CIMP+ patients.
The MMR associated genes identified with mRMR and IFS
Similarly, the MMR associated genes can be identified using mRMR and IFS methods. As a result, 18 genes were selected based on the IFS curve shown in Fig. 1C and listed in Table III. The leave one out cross validation Sn, Sp, ACC and MCC of these 18 genes were 0.922, 0.985, 0.976 and 0.902, respectively. As shown in Fig. 2C, the 18 MMR associated genes can cluster the MMR+ and MMR-patients into the right groups. CAB39L, H2AFJ, TGFBR2, MLH1, SEC22B, BRD3, FBXO21, FOXO3, INO80D were highly expressed in MMR-patients while EIF5A, RAPGEF6, LYG1, HNRNPL, MTA2, HPSE, STRN3, MIR3916, RAB12 were highly expressed in MMR+ patients.
The direct overlap between CIN associated genes, CIMP associated genes and MMR associated genes
As three major mechanisms of colorectal cancer, we would like to investigate whether there were overlaps between CIN associated genes, CIMP associated genes and MMR associated genes. The Venn diagram of CIN associated genes, CIMP associated genes and MMR associated genes were shown in Fig. 3. It can be seen that none genes were common in these three gene lists. The overlap between CIN and CIMP was ADGRG6, the common gene between CIN and MMR was TGFBR2 and the overlap between CIMP and MMR was MLH1. The references of ADGRG6 was limited and its functions were largely unknown. Interestingly, TGFBR2 has been reported as a candidate driver gene in MSI colorectal cancer (41) and the MMR patients usually also show varying degrees of CIN (8). TGFBR2 may be key of the association of CIN and MMR. The correlation of MLH1 methylation and MMR status has been reported (42) and it confirmed the association of CIMP and MMR.
The cross talk between CIN, CIMP and MMR
Since there is little overlap between the CIN associated genes, CIMP associated genes and MMR associated genes identified by mRMR and IFS, we would like to investigate whether they have common downstream genes. To verify this, we used the workflow shown in Fig. 4 to investigate the cross talk between CIN, CIMP and MMR. The key is step (C) which identifies the genes that the CIN, CIMP and MMR affects, i.e. the downstream genes of CIN, CIMP and MMR. To do so, first we mapped the CIN associated genes onto the network and then, expanded them using RWR network on the network. At last, by comparing with random permutations, the significant RWR expanded genes were identified as the downstream of CIN. Similarly, the downstream genes of CIMP and MMR can be identified.
The numbers of downstream genes of CIN, CIMP and MMR with permutation P-value <0.05 were 745, 709 and 807, respectively. Fig. 5 showed the overlap among CIN, CIMP and MMR and there were 236 common downstream genes of CIN, CIMP and MMR. These 236 genes were shown in Table IV. To statistically evaluate the significance of overlap, we calculated the odds ratio and P-value using R package Super Exact Test (43). The results were shown in Fig. 6. The odds ratio of overlap was 60.3 and the P-value was smaller than 1e-320.
Table IV.Common downstream genes of chromosome instability, CpG island methylator phenotype and mismatch repair. |
The biological functions of the overlapped genes were investigated by enriching them onto KEGG and GO. The enrichment results were summarized in Table V. It can be seen that the significantly enriched KEGG pathways with FDR (false discovery rate) <0.05 were: hsa00770 Pantothenate and CoA biosynthesis, hsa00785 Lipoic acid metabolism and hsa04514 Cell adhesion molecules (CAMs). Similarly, the most significantly enriched GO terms were: GO:0015937 coenzyme A biosynthetic process, GO:0015936 coenzyme A metabolic process, GO:0033866 nucleoside bisphosphate biosynthetic process, GO:0034030_ribonucleoside bisphosphate biosynthetic process and GO:0034033 purine nucleoside bisphosphate biosynthetic process. These results indicated that the CIN, CIMP and MMR all affect biosynthetic and metabolic process and pathway to accelerate the tumorgenesis. In clinic, the metabolic syndrome was found to be able to increase the risk of colorectal cancer (44). And in colorectal cancer cell, there are aberration of various metabolites, such as nucleotides, amino acids, tricarboxylic acid, carbohydrates, and pentose-phosphate (45).
Table V.Kyoto Encyclopedia of Genes and Genomes and Gene Ontology enrichments of common downstream genes of chromosome instability, CpG island methylator phenotype and mismatch repair. |
As a complex disease, the colorectal cancer can be caused by several different mechanisms. The three well-known one were CIN, CIMP and MMR. They were different but not exclusive. We investigated the genes that were associated with CIN, CIMP and MMR, separately using mRMR and IFS methods. Then by direct overlapping the CIN associated genes, CIMP associated genes and MMR associated genes, they share little common genes. Therefore, they were highly possible to interact with each other indirectly. To verify this idea, we identified the downstream genes that the CIN associated genes, CIMP associated genes and MMR associated genes may affect using RWR method. After the RWR analysis, the overlap between CIN, CIMP and MMR become significantly greater and the common downstream genes were involved in biosynthetic and metabolic process and pathway. These results can help explain the non-exclusiveness of CIN, CIMP and MMR and why they may co-occur from a protein-protein interaction network view. What's more, the common genes of CIN, CIMP and MMR can be possible targets of new broad-spectrum anti-cancer drugs that can treat more patients.
Acknowledgements
Not applicable.
Funding
The present study was supported by Health and Family Planning Commission of Zhejiang Province (grant no. 2013kYA212), National Natural Science Foundation of China (grant no. 31701151), Shanghai Sailing Program and The Youth Innovation Promotion Association of Chinese Academy of Sciences (CAS) (grant no. 2016245).
Availability of data and materials
The gene expression profiles of 585 colorectal cancer patients were obtained from GEO (Gene Expression Omnibus) with accession number of GSE39582.
Authors' contributions
RFW and TH designed the experiment. TMZ and TH performed the experiment, analyzed the data and wrote the manuscript.
Ethics approval and consent to participate
Not applicable.
Consent for publication
Not applicable.
Competing interests
The authors declare that they have no competing interests.
References
Siegel R, Desantis C and Jemal A: Colorectal cancer statistics, 2014. CA Cancer J Clin. 64:104–117. 2014. View Article : Google Scholar : PubMed/NCBI | |
Fearon ER and Vogelstein B: A genetic model for colorectal tumorigenesis. Cell. 61:759–767. 1990. View Article : Google Scholar : PubMed/NCBI | |
Li BQ, Huang T, Zhang J, Zhang N, Huang GH, Liu L and Cai YD: An ensemble prognostic model for colorectal cancer. PLoS One. 8:e634942013. View Article : Google Scholar : PubMed/NCBI | |
Jiang Y, Huang T, Chen L, Gao YF, Cai Y and Chou KC: Signal propagation in protein interaction network during colorectal cancer progression. BioMed Research International. 2013:2870192013. View Article : Google Scholar : PubMed/NCBI | |
Li BQ, Huang T, Liu L, Cai YD and Chou KC: Identification of colorectal cancer related genes with mRMR and shortest path in protein-protein interaction network. PLoS One. 7:e333932012. View Article : Google Scholar : PubMed/NCBI | |
Huang T, Li BQ and Cai YD: The integrative network of gene expression, microRNA, methylation and copy number variation in colon and rectal cancer. Curr Bioinform. 11:59–65. 2016. View Article : Google Scholar | |
Wu WK, Wang XJ, Cheng AS, Luo MX, Ng SS, To KF, Chan FK, Cho CH, Sung JJ and Yu J: Dysregulation and crosstalk of cellular signaling pathways in colon carcinogenesis. Crit Rev Oncol Hematol. 86:251–277. 2013. View Article : Google Scholar : PubMed/NCBI | |
Trautmann K, Terdiman JP, French AJ, Roydasgupta R, Sein N, Kakar S, Fridlyand J, Snijders AM, Albertson DG, Thibodeau SN and Waldman FM: Chromosomal instability in microsatellite-unstable and stable colon cancer. Clin Cancer Res. 12:6379–6385. 2006. View Article : Google Scholar : PubMed/NCBI | |
Walther A, Houlston R and Tomlinson I: Association between chromosomal instability and prognosis in colorectal cancer: A meta-analysis. Gut. 57:941–950. 2008. View Article : Google Scholar : PubMed/NCBI | |
Vedeld HM, Merok M, Jeanmougin M, Danielsen SA, Honne H, Presthus GK, Svindland A, Sjo OH, Hektoen M, Eknaes M, et al: CpG island methylator phenotype identifies high risk patients among microsatellite stable BRAF mutated colorectal cancers. Int J Cancer. 141:967–976. 2017. View Article : Google Scholar : PubMed/NCBI | |
Boland CR and Goel A: Microsatellite instability in colorectal cancer. Gastroenterology. 138:2073–2087.e3. 2010. View Article : Google Scholar : PubMed/NCBI | |
Guastadisegni C, Colafranceschi M, Ottini L and Dogliotti E: Microsatellite instability as a marker of prognosis and response to therapy: A meta-analysis of colorectal cancer survival data. Eur J Cancer. 46:2788–2798. 2010. View Article : Google Scholar : PubMed/NCBI | |
Marisa L, de Reyniès A, Duval A, Selves J, Gaub MP, Vescovo L, Etienne-Grimaldi MC, Schiappa R, Guenot D, Ayadi M, et al: Gene expression classification of colon cancer into molecular subtypes: Characterization, validation, and prognostic value. PLoS Med. 10:e10014532013. View Article : Google Scholar : PubMed/NCBI | |
Peng H, Long F and Ding C: Feature selection based on mutual information: Criteria of max-dependency, max-relevance, and min-redundancy. IEEE Trans Pattern Anal Mach Intell. 27:1226–1238. 2005. View Article : Google Scholar : PubMed/NCBI | |
Zhou Y, Zhang N, Li BQ, Huang T, Cai YD and Kong XY: A method to distinguish between lysine acetylation and lysine ubiquitination with feature selection and analysis. J Biomol Struct Dyn. 33:2479–2490. 2015. View Article : Google Scholar : PubMed/NCBI | |
Zhao TH, Jiang M, Huang T, Li BQ, Zhang N, Li HP and Cai YD: A novel method of predicting protein disordered regions based on sequence features. Biomed Res Int. 2013:4143272013. View Article : Google Scholar : PubMed/NCBI | |
Niu B, Huang G, Zheng L, Wang X, Chen F, Zhang Y and Huang T: Prediction of substrate-enzyme-product interaction based on molecular descriptors and physicochemical properties. Biomed Res Int. 2013:6742152013. View Article : Google Scholar : PubMed/NCBI | |
Zhang N, Wang M, Zhang P and Huang T: Classification of cancers based on copy number variation landscapes. Biochim Biophys Acta. 1860:2750–2755. 2016. View Article : Google Scholar : PubMed/NCBI | |
Liu L, Chen L, Zhang YH, Wei L, Cheng S, Kong X, Zheng M, Huang T and Cai YD: Analysis and prediction of drug-drug interaction by minimum redundancy maximum relevance and incremental feature selection. J Biomol Struct Dyn. 35:312–329. 2017. View Article : Google Scholar : PubMed/NCBI | |
Zhang N, Huang T and Cai YD: Discriminating between deleterious and neutral non-frameshifting indels based on protein interaction networks and hybrid properties. Mol Genet Genomics. 290:343–352. 2015. View Article : Google Scholar : PubMed/NCBI | |
Shu Y, Zhang N, Kong X, Huang T and Cai YD: Predicting A-to-I RNA editing by feature selection and random forest. PLoS One. 9:e1106072014. View Article : Google Scholar : PubMed/NCBI | |
Li BQ, You J, Huang T and Cai YD: Classification of non-small cell lung cancer based on copy number alterations. PLoS One. 9:e883002014. View Article : Google Scholar : PubMed/NCBI | |
Zhang PW, Chen L, Huang T, Zhang N, Kong XY and Cai YD: Classifying ten types of major cancers based on reverse phase protein array profiles. PLoS One. 10:e01231472015. View Article : Google Scholar : PubMed/NCBI | |
Huang T, Shu Y and Cai YD: Genetic differences among ethnic groups. BMC Genomics. 16:10932015. View Article : Google Scholar : PubMed/NCBI | |
Huang T, Wang M and Cai YD: Analysis of the preferences for splice codes across tissues. Protein Cell. 6:904–907. 2015. View Article : Google Scholar : PubMed/NCBI | |
Chen L, Zhang YH, Huang T and Cai YD: Identifying novel protein phenotype annotations by hybridizing protein-protein interactions and protein sequence similarities. Mol Genet Genomics. 291:913–934. 2016. View Article : Google Scholar : PubMed/NCBI | |
Li J, Chen L, Wang S, Zhang Y, Kong X, Huang T and Cai YD: A computational method using the random walk with restart algorithm for identifying novel epigenetic factors. Mol Genet Genomics. 293:293–301. 2018. View Article : Google Scholar : PubMed/NCBI | |
Li L, Wang Y, An L, Kong X and Huang T: A network-based method using a random walk with restart algorithm and screening tests to identify novel genes associated with Menière's disease. PLoS One. 12:e01825922017. View Article : Google Scholar : PubMed/NCBI | |
Chen L, Chu C, Kong X, Huang G, Huang T and Cai YD: A hybrid computational method for the discovery of novel reproduction-related genes. PLoS One. 10:e01170902015. View Article : Google Scholar : PubMed/NCBI | |
Szklarczyk D, Franceschini A, Wyder S, Forslund K, Heller D, Huerta-Cepas J, Simonovic M, Roth A, Santos A, Tsafou KP, et al: STRING v10: Protein-protein interaction networks, integrated over the tree of life. Nucleic Acids Res. 43:(Database Issue):. D447–D452. 2015. View Article : Google Scholar : PubMed/NCBI | |
Chen L, Zhang YH, Li J, Wang S, Zhang Y, Huang T and Cai YD: Deciphering the relationship between obesity and various diseases from a network perspective. Genes (Basel). 8(pii): E3922017. View Article : Google Scholar : PubMed/NCBI | |
Chen L, Pan H, Zhang YH, Feng K, Kong X, Huang T and Cai YD: Network-based method for identifying co-regeneration genes in bone, dentin, nerve and vessel tissues. Genes (Basel). 8(pii): E2522017. View Article : Google Scholar : PubMed/NCBI | |
Zhang J, Yang J, Huang T, Shu Y and Chen L: Identification of novel proliferative diabetic retinopathy related genes on protein-protein interaction network. Neurocomputing. 217:63–72. 2016. View Article : Google Scholar | |
Yang J, Huang T, Song WM, Petralia F, Mobbs CV, Zhang B, Zhao Y, Schadt EE, Zhu J and Tu Z: Discover the network mechanisms underlying the connections between aging and age-related diseases. Sci Rep. 6:325662016. View Article : Google Scholar : PubMed/NCBI | |
Chen L, Yang J, Huang T, Kong X, Lu L and Cai YD: Mining for novel tumor suppressor genes using a shortest path approach. J Biomol Struct Dyn. 34:664–675. 2016. View Article : Google Scholar : PubMed/NCBI | |
Chen L, Huang T, Zhang YH, Jiang Y, Zheng M and Cai YD: Identification of novel candidate drivers connecting different dysfunctional levels for lung adenocarcinoma using protein-protein interactions and a shortest path approach. Sci Rep. 6:298492016. View Article : Google Scholar : PubMed/NCBI | |
Niu B, Lu Y, Lu J, Chen F, Zhao T, Liu Z, Huang T and Zhang Y: Prediction of enzyme's family based on protein-protein interaction network. Current Bioinform. 10:16–21. 2015. View Article : Google Scholar | |
Chen L, Chu C, Lu J, Kong X, Huang T and Cai YD: A computational method for the identification of new candidate carcinogenic and non-carcinogenic chemicals. Mol Biosyst. 11:2541–2550. 2015. View Article : Google Scholar : PubMed/NCBI | |
Huang T, Liu CL, Li LL, Cai MH, Chen WZ, Xu YF, O'Reilly PF, Cai L and He L: A new method for identifying causal genes of schizophrenia and anti-tuberculosis drug-induced hepatotoxicity. Sci Rep. 6:325712016. View Article : Google Scholar : PubMed/NCBI | |
Hofree M, Shen JP, Carter H, Gross A and Ideker T: Network-based stratification of tumor mutations. Nat Methods. 10:1108–1115. 2013. View Article : Google Scholar : PubMed/NCBI | |
Alhopuro P, Sammalkorpi H, Niittymäki I, Biström M, Raitila A, Saharinen J, Nousiainen K, Lehtonen HJ, Heliövaara E, Puhakka J, et al: Candidate driver genes in microsatellite-unstable colorectal cancer. Int J Cancer. 130:1558–1566. 2012. View Article : Google Scholar : PubMed/NCBI | |
Parsons MT, Buchanan DD, Thompson B, Young JP and Spurdle AB: Correlation of tumour BRAF mutations and MLH1 methylation with germline mismatch repair (MMR) gene mutation status: A literature review assessing utility of tumour features for MMR variant classification. J Med Genet. 49:151–157. 2012. View Article : Google Scholar : PubMed/NCBI | |
Wang M, Zhao Y and Zhang B: Efficient test and visualization of multi-set intersections. Sci Rep. 5:169232015. View Article : Google Scholar : PubMed/NCBI | |
Forootan M, Tabatabaeefar M, Yahyaei M and Maghsoodi N: Metabolic syndrome and colorectal cancer: A cross-sectional survey. Asian Pac J Cancer Prev. 13:4999–5002. 2012. View Article : Google Scholar : PubMed/NCBI | |
Brown DG, Rao S, Weir TL, O'Malia J, Bazan M, Brown RJ and Ryan EP: Metabolomics and metabolic pathway networks from human colorectal cancers, adjacent mucosa, and stool. Cancer Metab. 4:112016. View Article : Google Scholar : PubMed/NCBI |