nomogram integrating genomics with clinicopathologic ..."colorectal cancer," "colon cancer,"...

Chromatin, Epigenetics, and RNA Regulation

Nomogram Integrating Genomics withClinicopathologic Features Improves PrognosisPrediction for Colorectal CancerYongfu Xiong1,Wenxian You1, Min Hou2, Linglong Peng1, He Zhou1, and Zhongxue Fu1

Abstract

The current tumor staging system is insufficient for predictingthe outcomes for patients with colorectal cancer because of itsphenotypic and genomic heterogeneity. Integrating gene ex-pression signatures with clinicopathologic factors may yield apredictive accuracy exceeding that of the currently available sys-tem. Twenty-seven signatures that used gene expression datato predict colorectal cancer prognosis were identified and re-analyzed using bioinformatic methods. Next, clinically annotat-ed colorectal cancer samples (n¼ 1710) with the correspondingexpression profiles, that predicted a patient's probability of can-cer recurrence, were pooled to evaluate their prognostic valuesand establish a clinicopathologic–genomic nomogram. Only 2of the 27 signatures evaluated showed a significant associationwith prognosis and provided a reasonable prediction accuracyin the pooled cohort (HR, 2.46; 95%CI, 1.183–5.132,P< 0.001;AUC, 60.83; HR, 2.33; 95% CI, 1.218–4.453, P < 0.001; AUC,

71.34). By integrating the above signatures with prognosticclinicopathologic features, a clinicopathologic–genomic nomo-gram was cautiously constructed. The nomogram successfullystratified colorectal cancer patients into three risk groups withremarkably different DFS rates and further stratified stage II andIII patients into distinct risk subgroups. Importantly, amongpatients receiving chemotherapy, the nomogram determinedthat those in the intermediate- (HR, 0.98; 95% CI, 0.255–0.679, P < 0.001) and high-risk (HR, 0.67; 95% CI, 0.469–0.957, P ¼ 0.028) groups had favorable responses.

Implications: These findings offer evidence that genomicdata provide independent and complementary progno-stic information, and incorporation of this information re-fines the prognosis of colorectal cancer. Mol Cancer Res; 16(9);1373–84. �2018 AACR.

IntroductionColorectal carcinoma is the third most commonly diagnosed

malignant disease and the second leading cause of cancer deathworldwide (1). Despite advances in colorectal cancer screening,diagnosis, and curative resection, its prognosis is not entirelysatisfactory because optimalmanagement and individual therapystrategies still present great challenges, as colorectal cancer is awell-recognizedheterogeneous disease. Currently, treatment deci-sions and prognoses for patients with colorectal cancer are pri-marily driven by assessment of the tumor–node–metastasis(TNM) staging system, which is based merely on anatomicalinformation (2). Previous studies revealed that patients with stageI colorectal cancer have a 5-year survival rate of approximately93%, which decreases to approximately 80% for patients withstage II disease and to 60% for patients with stage III disease (3).

However, discrepancies in the survival outcomes of patients at thesame stage and receiving similar treatments are commonlyobserved.

More importantly, according to the current TNM stage, adju-vant therapy is generally recommended for all patients withstage III disease (4). However, patients with T1–2N1M0 tumors(stage IIIA) have significantly higher survival rates than thosewith stage IIB tumors (3), suggesting that stratifying patients athigh risk of recurrence, who are most likely to benefit fromadjuvant therapy, is critical. Therefore, substantial effort has beenmade to discover new clinicopathologic indicators to optimizethe current staging system. To date, some clinicopathologic fea-tures, such as emergency presentation, adjacent organ involve-ment (T4), intestinal perforation or obstruction, high tumorgrade, inadequate sampling of lymph nodes, and lymphatic/vascular invasion, with prognostic value to classify colorectalcancer patients who are at a high risk of recurrence have beenapplied in clinical practice.However, these factors are all relativelyweak and insufficient to identify colorectal cancer patients whomay benefit from adjuvant therapy, which leads to potentialunder-treatment or over-treatment (5).

Colorectal cancer biological behavior (e.g., recurrence, metas-tasis, drug resistance) is a tightly regulated process that requiresthe aberrant expression of related genes to empower carcinomacells with their corresponding abilities. Accordingly, before clin-icopathologically detectable changes occur, the underlying altera-tions necessary for recurrence have already occurred in the pri-mary colorectal cancer, providing the possibility to developrobust prognostic tools by using multiple genes in combination

1Department of Gastrointestinal Surgery, The First Affiliated Hospital ofChongqing Medical University, Chongqing, China. 2Department of Oncology,Affiliated Hospital of North Sichuan Medical College, Nanchong, China.

Note: Supplementary data for this article are available at Molecular CancerResearch Online (http://mcr.aacrjournals.org/).

Y. Xiong, W. You, and M. Hou contributed equally to this article.

CorrespondingAuthor: Zhongxue Fu, The First Affiliated Hospital of ChongqingMedical University, Chongqing 400016, China. Phone: 28-8754-3627; E-mail:[email protected]

doi: 10.1158/1541-7786.MCR-18-0063

�2018 American Association for Cancer Research.

MolecularCancerResearch

www.aacrjournals.org 1373

on July 25, 2020. © 2018 American Association for Cancer Research. mcr.aacrjournals.org Downloaded from

Published OnlineFirst May 21, 2018; DOI: 10.1158/1541-7786.MCR-18-0063

http://crossmark.crossref.org/dialog/?doi=10.1158/1541-7786.MCR-18-0063&domain=pdf&date_stamp=2018-8-20http://mcr.aacrjournals.org/

(6). During the last decade, gene signatures have shown greatpromise in predicting the long-term outcomes and treatmentresponses of individual patients (7). Of note, because of theoutstanding ability to predict the prognoses of patients withbreast cancer, multigene assays, such as Oncotype DX andMammaPrint, have been successfully approved by the US Foodand Drug Administration and are available in routine clinicalpractice (8). Actually, genomic signatures to predict the progno-sis of colorectal cancer have also been continually developed inthe past 10 years, but none are commercially used in the clinic (9).For example, Agesen and colleagues established a 13-gene expres-sion classifier, ColoGuideEx, for prognosis predictions specificto patients with stage II colorectal cancer (10). Based on theessential role of lipid metabolism in carcinogenesis, Teodoro andcolleagues identified a group of 4 genes that predict survival inintermediate-stage colon cancer patients, allowing the delinea-tion of a high-risk group that may benefit from adjuvant therapy.In addition, Smith (11), Chen (12), Oh (13), and Popoviciand colleagues (14) also published their own signatures basedon different mechanisms. Despite a large number of studies,no signatures have remained credible in either meta-analyses orprospective trials (15). However, these published signaturesclearly show low prediction accuracies but moderate clinicalusefulness (15). Moreover, when confined to a specific colorectalcancer stage, promising results regarding risk stratification havealso been reported. Therefore, investigating the predictive abilityof these published signatures on comprehensive large-scale data-sets and identifying whether any can be used to clinically guidetreatment decisions is necessary.

Currently, while doubts about the predictive value of clinico-pathologic features are increasing, they still provide the mostreliable guidelines for the prognostication and treatment ofcolorectal cancer (16). Thus, we hypothesized that integratinggenomic signatures with clinicopathologic features in a modelwould yield a predictive accuracy exceeding that of the currentlyavailable prognostic system. Nomogram is a statistical predictionmodel that combines multiple prognostic factors to make intu-itive graphical and individualized predictions (17). Here, weaimed to apply a systematic approach to evaluate the clinicalusefulness of colorectal cancer–related signatures and then con-struct a composite clinicopathologic–genomic nomogram byintegrating factors with potential prognostic value in a trainingset. Moreover, using another independent set, the capacity of thenomogram to stratify colorectal cancer patients most likely tobenefit from chemotherapy was further validated.

Materials and MethodsPatients and prognostic signatures

To identify gene expression data arrayed using the Affymetrixplatform with clinically annotated data, we systematically search-ed Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/), ArrayExpress (http://www.ebi.ac.uk/arrayexpress/)and related literature with the keywords "colorectal cancer,""colorectal cancer," "colon cancer," "survival," "relapse,""recurrence," "prognostic," "prognosis," and "outcomes" pub-lished through August 1, 2017. For some datasets whose clinicaldata did not accompany their gene expression profiles, we eithersearched the supplements or contacted one or more of theinvestigators to obtain the missing information. Moreover, data-sets with small sample sizes and duplications were excluded. Raw

microarray data and the corresponding clinical data of thesedatasets were retrieved and manually organized when available.Only patients diagnosed with colorectal cancer having clinico-pathologic and survival information available were included.Patients with follow-up or survival times of less than 1 monthas well as patients with missing or insufficient data on age, localinvasion, lymph node metastasis, distant metastasis, and TNMstage were excluded from subsequent processing. Eventually, allpatients satisfying the inclusion criteria were combined andsummarized in Supplementary Tables S1 and S2.

Expression data processingFor raw CEL files available from Affymetrix microarrays, the

datawere normalized and annotated using aMAS5 algorithm andthe corresponding annotationfiles fromRBioconductor to obtainsummarized values for each probe set; otherwise, preprocesseddata as provided by the contributorswere used. For each sample inevery data set, measurements without a gene annotation wereexcluded, and multiple probe sets corresponding to a single genewere summarized into a gene symbol by taking the most variableprobe set measured by the interquartile range (IQR).

Identification and analysis of gene signatures potentiallyrelated to colorectal cancer prognosis

Gene signatures potentially related to colorectal cancer prog-nosis were systematically retrieved from PubMed. The search wasrestricted to recent papers to increase validity (from January 2004to August 2017). The selection criteria are detailed in Supple-mentary Fig. S1. Articles that provided a list of differentiallyexpressed genes in primary tumor samples associated with colo-rectal cancer prognosis were included in our study. Studies basedon tissue microarray and those that were exclusively focused ondifferences between stages or between primary tumors andmetas-taseswere excluded. The signaturesfinally included inour analysisare described in Table 1 (detailed descriptions provided in Sup-plementary Table S3). In addition, the probe sets or genes of thosesignatures were re-annotated using the SOURCE web tool toaddress the retired gene symbols and their differences in thetested platforms. The re-annotated genes were then subjected tobiological function enrichment analysis, and the online analyticaltool DAVID (Database for Annotation, Visualization and Inte-grated Discovery; ref. 18) was used to enrich gene ontology (GO)functions and Kyoto Encyclopedia of Genes and Genomes(KEGG) pathways. GO terms and KEGG pathways with signifi-cant enrichment false discovery rate (FDR) values less than 0.05were selected for further analysis. In addition, genes from theabovementioned signatures were mapped and imported into theRetrieval of Interacting Genes/Proteins (STRING) 9.1 database,which queried the human protein–protein interaction networkfor interactions between effective linkers and seeds to construct afunctional subnetwork. Cluster analyses were performed usingcorrelation distance metrics and the average linkage agglomera-tion algorithm (R package nclust version 1.9.0). Nonnegativematrix factorization (NMF) was done with the NMF package(version 0.20.5) and standard strategies.

Subclass prediction of colorectal cancer patients in thetraining set

The re-preprocessed microarray dataset, which represents thegenomic features of the individual, was classified with the prog-nostic signatures identified above using the nearest template

Xiong et al.

Mol Cancer Res; 16(9) September 2018 Molecular Cancer Research1374



http://www.ncbi.nlm.nih.gov/geo/http://www.ncbi.nlm.nih.gov/geo/http://www.ebi.ac.uk/arrayexpress/http://mcr.aacrjournals.org/

prediction (NTP) method (19) as implemented in Gene Patternsoftware (Broad Institute of Harvard and MIT, Boston, MA;ref. 20). NTP requires only a list of prespecified template signaturegenes and a dataset to be tested and not a corresponding trainingdataset to capture good and poor gene expression patterns in eachsample. Briefly, a template containing representative expressionpatterns of the signature genes was defined based on publishedgene signatures from their respective studies. The proximity of thesignature gene expression patterns of the sample to the templatewas evaluated by calculating the cosine distance. The FDR wasused to correct the P values for multiple hypothesis testing, andprediction analysis was performed separately for each dataset. Aprediction of good or poor for the related signature was deter-mined based on FDR predictions

published as valid prognostic tools in colorectal cancer wereobtained from 507 potentially related articles. Among thesesignatures, those with genes not clearly described in theirrespective studies or investigating the prognostic value basedon tissue array, immune cells, circulation blood, or RT-PCRresults were further excluded (see flowchart in SupplementaryFig. S1 for details). Finally, a total of 27 signatures from26 studies met all of the inclusion criteria and were retainedfor subsequent analysis (Supplementary Table S3). Amongthese signatures, almost all were obtained from experimentsbased on colorectal cancer tissues except for one, which wasderived from an animal experiment (23). The 27 signaturescontained 1274 total genes, among which 1041 were unique,and the signature sizes ranged from 4 to 264 (Fig. 1A). Thetop overlapping genes in the above signatures were FN1 (10times), CYP1B1 and POSTN (6 times), 7 genes (5 times), 7

genes (4 times), 29 genes (3 times), and 107 genes (twice). As isclearly shown in Fig. 1A, none of these genes appeared inall signatures. In addition, some repeated genes in differentsignatures even presented opposite prognostic values. Thislower-than-expected overlap, at least in part, accountedfor the low reproducibility of the signatures and the failure inclinical practice.

Accumulating evidence strongly suggests that the prognosisof colorectal cancer patients, such as recurrence and metastasis,is tightly regulated by aberrantly expressed genes via specificbiological processes (16). To gain insight into the underlyingbiological meanings of the gene set obtained from the abovesignatures, enrichment (GO and KEGG) and protein–proteininteraction (PPI) analyses were conducted (Supplemen-tary Table S4). Figure 1B–D shows that specific GO cate-gories closely related to colorectal cancer prognosis, such as

Figure 1.

Bioinformatic analysis of gene lists included in this study. A, Landscape distribution characteristics of the gene lists as reported in 27 prognostic signatures.Genes with good and poor prognostic values in each signature are shown as short green and red vertical bars, respectively. The number of gene occurrences isgraphically depicted on the top. Each row represents one signature annotated by the first author of the corresponding article. All the unique genes weresorted alphabetically. B–E, GO and KEGG pathway analyses of genes in the 27 signatures. The vertical axis represents Go or KEGG pathway annotations. Thehorizontal axis represents the number of genes assigned to the corresponding annotation. B, Biological process; C, cellular components; D, molecular function;and E, KEGG pathway. F, Protein–protein interaction (PPI) networks. Nodes represent the proteins, and edges represent the physical interaction. The blackand gray nodes represent geneswith good and poor prognostic values, respectively. Colorectal cancer recurrence-related pathways that were significantly enrichedby KEGG analysis form some closely connected regions and are highlighted in the colored box.

Xiong et al.




http://mcr.aacrjournals.org/

cell cycle (FDR < 0.0001), regulation of cell proliferation(FDR < 0.0001), extracellular matrix (FDR < 0.0001), andchemokine activity (FDR < 0.0001), were significantly enriched.Additionally, some KEGGpathways well known to be involved incolorectal cancer metastasis and recurrence were enrich-ed, including colorectal cancer metastasis signaling (FDR <0.0001), wnt-b catenin signaling (FDR ¼ 0.0002), VEGFsignaling (FDR ¼ 0.0008), and chemokine signaling (FDR ¼0.0021; Fig. 1E). More importantly, a visualized network of thesegene interactions revealed that the abovementioned pathways notonly form highly integrated modules but also play essential rolesin the entire network (Fig. 1F). These findings suggested thatnotwithstanding the low reproducibility in the system test andclinical validation, genes obtained from colorectal cancer signa-tures actually have potential prognostic value.

Furthermore, to assess the global performance of thesesignature and construct a composite nomogram, publicly avail-able datasets with whole-genome profiles and correspondingclinical information in colorectal cancer patients were screenedand downloaded (see flowchart in Supplementary Fig. S1 fordetails). Ultimately, 1710 colorectal cancer patients pertainingto 9 cohorts (Supplementary Table S5) met the set criteria andwere retained. Although the profiles were obtained using dif-ferent technologies and based on fresh or formalin-fixed andparaffin-embedded (FFPE) specimens as source material, theyshowed no evidence of cohort-bias clustering by principalcomponent analysis (PCA; Supplementary Fig. S2A–S2C) orhierarchical clustering (Supplementary Fig. S2D), suggestingthat patients from different cohorts in the present study couldbe blended together. Moreover, to balance the baseline char-acteristics, all eligible patients were randomly divided into atraining set (n ¼ 855) and a validation set (n ¼ 855). Thedemographic and major clinicopathologic characteristics of thepatients are summarized in Supplementary Table S2. As weexpected, no significant differences were observed between thestatistical properties of the training and validation sets.

Global prognostic performance of the published signaturesTo determine the correlations of signatures and colorectal

cancer patient outcomes, the prognostic performances of the27 signatures were evaluated in the training set using a modifiedNTP method as previously described (19). Among the 27 signa-tures evaluated, all except one (Teodoro's (24) signature had theonly poor prognostic gene) were able to confidently stratifypatients (FDR < 0.05) into good and poor subgroups. Table 1and Fig. 2A summarize the prediction results obtained for each ofthe 855 patients. Popovici signature (34) was the most prevalentprediction in the training set (83.6%; 615of 855), whereas Agesensignature (42) was securely identified in only 19.9% (167 of 855)of colorectal cancer patients. Of note, conflicting prognostic out-comeswere commonly observed among the training set, and only34% patients had consistent results (good or poor prognosis) inmore than 5 signatures. However, as demonstrated in Fig. 2A, theoverall tendency of the NTP analysis was consistent and roughlyconsistent with previous studies (25). In addition, global CramerV coefficients were calculated for each signature to gain insightinto their concordances. Unsupervised clustering based on thesecoefficients indicated a substantial association among these sig-natures (Fig. 2B). Obviously, 5 signatures obtained from thestudies of Oh (13), Minh (11), Chang (9), Jiang (26), andFritzmann and colleagues (27)were at leastmoderately correlated

with each other. Moreover, another 4 signatures (Toshiaki's,Aziz's, Chen's, Xu's) were also clustered with similar coefficients.These findings are largely consistent with the initial purpose ofeach signature, which further confirms the potential prognosticvalue of these signatures.

We then applied Cox regression analysis to assess whether anysignatures were statistically significant for colorectal cancer–related recurrence/progression and independent of clinical fea-tures in the training set. The univariate and clinical factor-adjustedmultivariate analysis results are displayed in SupplementaryTable S6 and Fig. 2C. Univariate analyses revealed that age(HR,1.12; 95% CI, 1.021–1.269, P ¼ 0.048), local invasion(T stage; HR, 2.21; 95% CI, 1.139–7.029, P ¼ 0.027), lymphnode metastasis (N stage; HR, 2.28; 95% CI, 1.189–3.784, P ¼0.021), distant metastasis (M stage; HR, 2.46; 95% CI, 1.363–4.453, P ¼ 0.003), and TNM stage (HR, 2.15; 95% CI, 1.165–3.971, P ¼ 0.014) were confidently associated with poor prog-nosis. However, with respect to genomic factors, only 8 signatureswere validated to have significant effects on DFS (Fig. 2C). Next,we used multivariate Cox analysis with the covariates, includingage, local invasion, lymph node metastasis, distant metastasis,and TNM stage, to evaluate the predictive values of the 8 signa-tures. As shown in Fig. 4C, after each covariate was adjusted,Popovici signature (HR, 1.92; 95% CI, 1.298–2.841, P ¼ 0.001)and Fritzmann signature (HR, 2.16; 95% CI, 1.144–4.069,P < 0.0001) remained powerful enough to indicate prognosis inthe training set for DFS, which revealed that these two signaturesare independent prognostic factors.

Construction and comparison analysis of the compositenomogram

Accurately predicting DFS in patients with colorectal cancer isimportant for counseling, follow-up planning, and selection ofappropriate adjuvant therapy. Thus, toprovide the clinicianwith aquantitative and user-friendly method to generate individualizedpredictions, we constructed a nomogram that integrated thegenomic signatures identified above with clinicopathologic fea-tures (Fig. 3A). Calibration plots revealed excellent agreementbetween the nomogram-predicted probabilities and the actualobservations of 1-, 3-, and 5-year DFS (Fig. 3B). The nomogramdemonstrated that distant metastasis made the largest contribu-tion to prognosis, followed by the TNM stage and lymph nodemetastasis. Age and Popovici signature had moderate effects onsurvival rate. Each category within these variables was assigned apoint on the top scale based on the coefficients from Cox regres-sion. By summing all of the assigned points for the seven variablesand drawing a vertical line between total points and the survivalprobability axis, we easily obtained the estimated probabilities of1-, 3-, and 5-year DFS. The risk score cutoff values (�10.6, 10.6–21.2, and �21.2) were selected in terms of total points to stratifypatients into roughly equal tertiles in the training set, whichaccurately divided patients into low- (reference), intermediate-(HR, 2.227; 95% CI, 1.507–3.291, P < 0.001), and high-risk(HR, 6.787; 95% CI, 4.786–9.624, P < 0.001) subgroups withsignificantly different DFS rates (Fig. 3C and D).

Although the TNM stage in combination with other clinicalfactors is a well-recognized prediction system for colorectal cancerprognosis (28), its effectiveness urgently needs to be increased. Todetermine whether the genomic signatures added additionalprognostic value to the current system, time-dependent receiveroperating characteristic (ROC) analysis was applied to compare

A Nomogram to Predict Colorectal Cancer Prognosis

www.aacrjournals.org Mol Cancer Res; 16(9) September 2018 1377




the performances between the nomogram, clinicopathologic fac-tors, and genomic signatures (Fig. 3E). Expectedly, the nomogramachieved the greatest area under the ROC curve (AUCs at 5-yearDFS, 78.07, 71.40, 71.34, and 60.83, P < 0.05), suggesting that theintegrated clinicopathologic genomic nomogram had a prognos-tic performance superior to those of clinicopathologic and geno-mic information by themselves.

Validating the nomogram for stratifying colorectal cancerpatient risks

To validate the correlation between the nomogram score (totalpoints) and DFS statuses of patients with colorectal cancer,

Kaplan–Meier analysis and log-rank tests based on the samecutoff value were conducted on the validation set. As shownin Fig. 4A and B, applying the clinicopathologic–genomic nomo-gram stratified patients into three distinct risk subgroups withsignificantly different DFS rates. In addition, the prognostic accu-racy of the nomogram was remarkably better than those of bothclinicopathologic and genomic information by themselves (Fig.4C), which was consistent with the results obtained from thetraining set. Disagreements on potentially appropriate candidatesfor adjuvant therapy prevail, especially for patients belonging toAJCC stages II and III. In the present study, the survival ratespredicted by the nomogram were significantly distinct between

Figure 2.

Global prognostic performance of the published colorectal cancer signatures. A, Concordance of signature-based prediction results in the training set. Eachcolumn represents the prediction of each individual sample. The orange, blue, and white bars indicate good, poor, and uncertainty prognoses of thecorresponding signature, respectively. B, Heatmap of Cramer V coefficients showing correlation between these published signatures. C, Analysis of prognosticperformance using Cox regression; left: univariate analysis; right: clinical factors adjusted for multivariate analysis.

Xiong et al.





the Kaplan–Meier curves (Fig. 4D and E) in patients categorizedby the abovementioned major clinicopathologic features (AJCCstages II and III).

Moreover, to ensure the effectiveness of statistical power,patients with available chemotherapy information were integrat-ed together regardless of the set to which they belonged (trainingset or validation set). Intriguingly, we noted that adjuvant che-motherapy did not enhanceOS in all 869 patients (HR, 0.98; 95%CI, 0.769–1.258, P ¼ 0.873; Fig. 5A) or in patients who weredefined as low risk by the nomogram (HR, 0.93; 95% CI, 0.515–1.695, P ¼ 0.824; Fig. 5B). However, patients in the nomogramdefined as intermediate risk (HR, 0.98; 95% CI, 0.255–0.679,P < 0.001; Fig. 5C) or high risk (HR, 0.67; 95% CI, 0.469–0.957,P ¼ 0.028; Fig. 5D) had favorable responses to adjuvant chemo-therapy. Taken together, these results suggested that the clinico-pathologic–genomic nomogram has the discriminatory powerto identify colorectal cancer patients who are suitable candidatesfor adjuvant chemotherapy.

DiscussionAlthough surgery remains the mainstay of curative treatment,

the prognosis of patients with colorectal cancer, especially forlong-term outcome, depends more on pre- or postoperativeindividualized therapies (29), including various strategies ofchemotherapy and chemoradiation. However, as a well-recog-nized heterogeneous cancer, its prognosis varies significantly, andoptimal management presents challenges. Currently, adjuvantchemotherapy is regarded as standard treatment for patients withstage III colon cancer (4). However, patients with T1–2N1M0tumors (stage IIIA) have significantly higher survival rates thanthose with stage IIB tumors (3). Moreover, for some patients,long-term survival continues to be jeopardized by the high risk ofrecurrence even after chemotherapy, and personalized therapyneeds to further optimized. With respect to stage II patients,disagreement regarding which patients are potentially appropri-ate candidates for adjuvant therapy prevails. The MOSAIC trial

Figure 3.

Development of the composite clinicopathologic–genomic nomogram. A, Composite nomogram to predict disease-free survival (DFS) for patients with colorectalcancer. Each category within these variables was assigned to a point on the top scale based on the coefficients from Cox regression. By summing all of theassigned points for the seven variables and drawing a vertical line between the total points and survival probability axis, we easily obtained the estimatedprobabilities of 1-, 3-, and 5-year DFS. B, Calibration curves for predicting patient survival at 1, 3, and 5 years. The actual DFS is plotted on the y-axis, andthe nomogram-predicted DFS is plotted on the x-axis; a plot along the 45-degree line would indicate a perfect calibration model in which the predictedprobabilities are identical to the actual outcomes. C, The total point distribution of each colorectal cancer patient and their disease-free survival time and status.The blue dotted line represents the cutoff value dividing patients into low-, intermediate-, and high-risk groups. D, Kaplan–Meier survival curves of DFS fordifferent risk groups using the nomogram in the training set. E, Time-dependent ROC curves comparing the prognostic accuracies of 5-year disease-freesurvival among the prognostic signatures combined with clinicopathologic features and the nomogram in the training set. The AUC 95% CIs were calculatedfrom 1,000 bootstraps of the survival data. ROC, receiver operator characteristic. AUC, area under the curve.






(30) has shown only a 20% benefit from adjuvant therapy at 3years for patients with stage II disease and a 28% benefit forpatients with recognized clinical risk factors. Thus, for approxi-mately 75% of stage II patients who receive curative surgery, theadministration of adjuvant chemotherapy is unnecessary andharmful. Unfortunately, low-risk stage II patients have not beenaccurately defined. In addition, patients with microsatellite insta-bility (MSI colorectal cancer) have a better prognosis (31) and donot benefit from 5-fluorouracil (5-FU)–based adjuvant chemo-therapy (32), similar to patients with CpG island methylation(33). All the abovementioned studies highlight that the rationalapplication of therapeutic strategies requires accurate prognosisprediction and risk stratification for colorectal cancer patients.

Considering the inherent deficiency of TNM stages, substantialeffort has been placed on their optimization (2), and they havebeen continuously developed over the past decade. Substantialchanges in colorectal cancer classifications, based mainly on theanalysis of Surveillance, Epidemiology, and End Results (SEER)data, were made in the seventh edition (34). However, a com-

parative studydemonstrated that the seventh TNMeditiondidnotprovide a greater accuracy for predicting colorectal cancer patientprognoses but rather resulted in amore complex classification fordaily clinical use (35). The eighth edition of the colorectal cancerstaging system, containing major advances, including the intro-duction of molecular markers, such as MSI, KRAS, and BRAFmutations, to strengthen its discriminatory power, were imple-mented worldwide on January 1, 2018 (2). Nonetheless, theforthcoming edition may not provide substantial improvementbecause the prognostic value of clinicopathologic factors haspeaked. Moreover, the current models still lack genomic infor-mation, which may directly reflect the heterogeneity of colorectalcancer and present substantial potential prognostic value.

Genomic signatures predicting the prognosis of colorectalcancer have been continually developed in the past 10 years. Inthe present study, we retrieved 81 articles related to the develop-ment of signatures for predicting colorectal cancer prognosis,including OS, DFS, disease-specific survival (DSS), and varioustypes of recurrence. For instance, based on the association

Figure 4.

Verifying the prognostic value of the nomogram in the validation set. A, The total point distribution of each colorectal cancer patient and their disease-freesurvival time and status. The blue dotted line represents the cutoff value dividing patients into low-, intermediate-, and high-risk groups. B, Kaplan–Meiersurvival curves of DFS for different risk groups using the nomogram in the validation set. C, Time-dependent ROC curves comparing the prognostic accuraciesof 5-year disease-free survival among the prognostic signatures in combination with clinicopathologic features and the nomogram in the validation set.D and E, Kaplan–Meier survival curves for patients in different risk subgroups, which were stratified by TNM stages II (D) and III (E).

Xiong et al.





between the aberrantly expressed gene and the duration ofindividual DFS, Zhang and colleagues (36) constructed a sixRNA-based classifier, which was further validated to have reliablepredictive power for detecting recurrence in patients with stage IIcolon cancer. Similarly, Anna and colleagues (23) developed agenomic profile specific for adult intestinal stem cells (ISC),whichwas highly enriched in genes that positively predicted the risk ofrecurrence. Patients bearing primary colorectal cancers with highaverage expression levels of ISC genes had a relative risk of relapsethat was 10-fold higher than that for patients with low levels (23).Moreover, recent evidence (37) suggests that gene signatures thatclosely reflect specific biological processes or oncogenic pathwaystatuses also have potential prognostic values for stratifyingcolorectal cancer patients. Even though almost all of these sig-natures were proven to have prognostic value in their respectivepublications, none are routinely used in clinical practice.

Potential explanations for the unsatisfactory results must beconsidered. Thus, in the current study, we exhaustively reviewedand analyzed the published multigene signatures in colorectalcancer. To eliminate the potential confounding effects, only dataderived from the Affymetrix human genome platform havingclearly described genes were included. Of the 81 publishedsignatures, 27 signatures from 26 publications met all of theinclusion criteria and were retained (Table 1; SupplementaryTable S3). The 27 signatures contained 1,274 total genes, among

which 1,041were unique, and the signature sizes ranged from4 to264 (Fig. 1A). The top overlapping genes in the above signatureswere FN1 (10 times), CYP1B1 and POSTN (6 times), 7 genes(5 times), 7 genes (4 times), 29 genes (3 times), and 107 genes(twice). As is clearly shown in Fig. 2A, none of these genesappeared in all signatures. These results were expected becausethey were previously reported (15) and in agreement with thesystematic analyses of lung cancer and breast cancer prognosticsignatures (38). Obviously, the slight overlap of these signaturesin gene identity was perplexing but sufficiently explained the lowreproducibility. Moreover, genetic heterogeneity was recentlyshown to exist not only between but also within tumors,meaningthat biological statuses, including oncogenic and carcinostaticstates, might regularly vary in the same patient (39). Our NTPanalysis revealed that some patients concurrently harbored morethan one signature, which further confirmed the abovementionedresults and indicated that their tumors were genomic instability atthe individual level. Taken together, the different signaturesmightreflect diverse biological processes and be active in different oridentical individual tumors, which partially accounts for thewide nonoverlapping among signatures constructed in differentcolorectal cancer studies.

In terms of global performance, in the training set, 8 of 27signatures showed a significant association with prognosisand could reasonably predict the outcome. However, after

Figure 5.

Effect of chemotherapy on overallsurvival in patients stratified by thenomogram. A, Kaplan–Meier survivalcurves for all patients who receivedchemotherapy. B–D, Effects ofadjuvant chemotherapy on overallsurvival in the different subgroups.Low-risk group (B), intermediate-riskgroup (C), and high-risk group (D).






multivariate adjustment, only 2 signatures remained powerfulenough to indicate prognosis for DFS, which directly demon-strated the low reproducibility of colorectal cancer genomicsignatures. More importantly, Popovici signature, Fritzmannsignature and TNM stage had remarkable prognostic abilitiesand were independent of each other, indicating that thesesignatures may be used to refine the current prognostic modeland facilitate further stratification of patients with colorectalcancer at the same TNM stage.

Our primary goal was to construct a nomogram that couldintegrate genomic signatures with clinicopathologic variables in alarge cohort of colorectal cancer patients to add additional prog-nostic value to the current staging system. Our multivariateanalysis in the training set revealed that age, local invasion, lymphnode metastasis, distant metastasis, TNM stage, Popovici signa-ture andFritzmann signaturewere independent prognostic factors(Supplementary Table S6), which was highly consistent withprevious studies on colorectal cancer risk factors (40, 41). Accord-ingly, we cautiously built a clinicopathologic–genomic nomo-gram using the abovementioned significant factors. The AUC ofthe nomogram showed a superior prediction accuracy comparedwith that of the combined clinical factors in the training andvalidation sets (Figs. 3E and 4C). Additionally, the calibrationplots showed optimal agreement between the expected andobserved survival probabilities (Fig. 3B), ensuring the reliabilityof our nomogram. Moreover, the nomogram successfully strati-fied colorectal cancer patients into three risk groups with remark-ably different DFS values based on total points, which was furtherconfirmed using the same cutoff value in the validation set.

Intriguingly, the TNMstaging system is themost important toolfor guiding treatment in clinical practice. However, disagreementprevails on potentially appropriate candidates for adjuvant ther-apy, especially for stage II and III colorectal cancer patients.Therefore, the identification of a high-risk subgroup among thesepatients is of urgent clinical needed. Our findings suggest that theclinicopathologic–genomic nomogramcould be a promising toolto stratify stage II and III patients into distinct risk subgroup andguide individualized therapy. Moreover, in the current study, weclearly revealed that chemotherapy provides a survival benefit topatients classified as intermediate- and high risk by the nomo-gram (Fig. 5).

Despite the promising results, there were several shortcomingsin this work. First, the NTP method uses only a list of signaturegenes to make class predictions in each patient's expression data,allowing the method to be less sensitive to differences in exper-imental and analytical conditions and applicable to every patientwithout optimization of the analysis parameters. In the currentstudy, we applied the NTP method for all prediction models,

which generally consisted of two major components: a genesignature and a prediction algorithm. Obviously, the intrinsiclimitation of the NTP method is that it inevitably causes thealgorithm lose its prognostic value. Second, the forthcoming 8thedition of the colorectal cancer TNM staging system may betterpredict prognosis, but data used in the present studywere from thecurrent 7th edition or the 6th edition. Third, the prognosticsignatures of colorectal cancer included in our study were notexhaustive, and some promising signatures were excludedbecause of insufficient information. In addition, some clinico-pathologic factors, even those well recognized to have prognosticvalue, were not included in the composite nomogram because ofthe low availability of information.

In summary, based on analyzing a large-scale colorectal cancercohort and published multigene prognostic signatures, we con-firmed herein for the first time that genomic information incombination with clinical data can improve patient prognosticevaluations and should be considered as an independent andcomplementary approach to the current clinicopathologic prog-nostic model. Furthermore, considering the wide promise of ournomogram, which integrates genomic signatures with clinico-pathologic features to improve the prognosis prediction of colo-rectal cancer, further systematic research is needed.

Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.

Authors' ContributionsConception and design: Y. Xiong, Z. FuDevelopment of methodology: W. YouAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): Y. Xiong, M. HouAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): Y. Xiong, M. Hou, L. Peng, H. ZhouWriting, review, and/or revision of the manuscript: Y. Xiong, W. You, L. PengAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): Z. Fu

AcknowledgmentsThis work was financially supported by a grant from the National Natural

Science Foundation of China (grant no. 81572319; project recipient: Z.. Fu).

The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely to indicatethis fact.

Received January 19, 2018; revised April 5, 2018; accepted May 2, 2018;published first May 21, 2018.

References1. Favoriti P, Carbone G, Greco M, Pirozzi F, Pirozzi RE, Corcione F. Worldwide

burden of colorectal cancer: a review. Updates Surg 2016;68:7–11.2. Amin MB, Greene FL, Edge SB, Compton CC, Gershenwald JE, Brookland

RK, et al. The Eighth Edition AJCC Cancer Staging Manual: Continuing tobuild a bridge fromapopulation-based to amore "personalized" approachto cancer staging. CA Cancer J Clin 2017;67:93–9.

3. O'Connell JB, Maggard MA, Ko CY. Colon cancer survival rates with thenew American Joint Committee on Cancer sixth edition staging. J NatlCancer Inst 2004;96:1420–5.

4. Mahar AL, Compton C, Halabi S, Hess KR, Weiser MR, Groome PA.Personalizing prognosis in colorectal cancer: A systematic review of the

quality and nature of clinical prognostic tools for survival outcomes. J SurgOncol 2017;116:969–82.

5. Dotan E, Cohen SJ. Challenges in themanagement of stage II colon cancer.Semin Oncol 2011;38:511–20.

6. Stintzing S, Stremitzer S, Sebio A, Lenz HJ. Predictive and prognosticmarkers in the treatment of metastatic colorectal cancer (mCRC): person-alized medicine at work. Hematol Oncol Clin North Am 2015;29:43–60.

7. Lu AT, Salpeter SR, Reeve AE, Eschrich S, Johnston PG, Barrier AJ, et al. Geneexpression profiles as predictors of poor outcomes in stage II colorectalcancer: a systematic review and meta-analysis. Clin Colorectal Cancer2009;8:207–14.

Xiong et al.





8. Xin L, Liu Y-H, Martin TA, Jiang WG. The era of multigene panels comes?The clinical utility of Oncotype DX and MammaPrint. World J Oncol2017;8:34–40.

9. Chang W, Gao X, Han Y, Du Y, Liu Q, Wang L, et al. Gene expressionprofiling-derived immunohistochemistry signature with high prognosticvalue in colorectal carcinoma. Gut 2014;63:1457–67.

10. Agesen TH, Sveen A, MerokMA, Lind GE, Nesbakken A, Skotheim RI, et al.ColoGuideEx: a robust gene classifier specific for stage II colorectal cancerprognosis. Gut 2012;61:1560–7.

11. Smith JJ, Deane NG, Wu F, Merchant NB, Zhang B, Jiang A, et al.Experimentally derived metastasis gene expression profile predictsrecurrence and death in patients with colon cancer. Gastroenterology2010;138:958–68.

12. Chen H, Sun X, Ge W, Qian Y, Bai R, Zheng S. A seven-gene signaturepredicts overall survival of patients with colorectal cancer. Oncotarget2016;8:95054–65.

13. Oh SC, Park YY, Park ES, Lim JY, Kim SM, Kim SB, et al. Prognostic geneexpression signature associated with two molecularly distinct subtypes ofcolorectal cancer. Gut 2012;61:1291–8.

14. Popovici V, Budinska E, Tejpar S, Weinrich S, Estrella H, Hodgson G, et al.Identification of a poor-prognosis BRAF-mutant-like population ofpatients with colon cancer. J Clin Oncol 2012;30:1288–95.

15. Phillips WA, Sanz-Pamplona R, Berenguer A, Cordero D, RiccadonnaS, Sol�e X, et al. Clinical value of prognosis gene expression signa-tures in colorectal cancer: a systematic review. PLoS One 2012;7:e48877.

16. Zhang ZY, Gao W, Luo QF, Yin XW, Basnet S, Dai ZL, et al. A nomogramimproves AJCC stages for colorectal cancers by introducing CEA, mod-ified lymph node ratio and negative lymph node count. Sci Rep2016;6:39028.

17. Iasonos A, Schrag D, Raj GV, Panageas KS. How to build and interpret anomogram for cancer prognosis. J Clin Oncol 2008;26:1364–70.

18. Dennis G Jr, Sherman BT, Hosack DA, Yang J, Gao W, Lane HC, et al.DAVID: database for annotation, visualization, and integrated discovery.Genome Biol 2003;4:P3.

19. Hoshida Y. Nearest template prediction: a single-sample-based flexibleclass prediction with confidence assessment. PLoS One 2010;5:e15543.

20. Reich M, Liefeld T, Gould J, Lerner J, Tamayo P, Mesirov JP. GenePattern2.0. Nat Genet 2006;38:500–1.

21. Fan C, Oh D, Wessels L, Weigelt B, Nuyten D, Nobel A, et al. Concordanceamong gene-expression-based predictors for breast cancer. N Engl J Med2006;355:560–9.

22. Guyot P, Ades AE, Ouwens MJ, Welton NJ. Enhanced secondary analysisof survival data: reconstructing the data from published Kaplan-Meiersurvival curves. BMC Med Res Method 2012;12:9.

23. Merlos-Suarez A, Barriga FM, Jung P, Iglesias M, Cespedes MV,Rossell D, et al. The intestinal stem cell signature identifies colorectalcancer stem cells and predicts disease relapse. Cell Stem Cell 2011;8:511–24.

24. Vargas T, Moreno-Rubio J, Herranz J, Cejas P, Molina S, Gonz�alez-VallinasM, et al. ColoLipidGene: signature of lipid metabolism-related genes topredict prognosis in stage-II colon cancer patients. Oncotarget 2015;6:7348–63.

25. Wu J, Zhou L, Huang L, Gu J, Li S, Liu B, et al. Nomogram integrating geneexpression signatures with clinicopathological features to predict survivalin operable NSCLC: a pooled analysis of 2164 patients. J Exp Clin CancerRes 2017;36:4.

26. Jiang Y, CaseyG, Lavery IC, Zhang Y, TalantovD,Martin-McGreevyM, et al.Development of a clinically feasible molecular assay to predict recurrenceof stage II colon cancer. J Mol Diagn 2008;10:346–54.

27. Fritzmann J, Morkel M, Besser D, Budczies J, Kosel F, Brembeck FH, et al. Acolorectal cancer expression profile that includes transforming growthfactor b inhibitor BAMBI predicts metastatic potential. Gastroenterology2009;137:165–75.

28. Kazem MA, Khan AU, Selvasekar CR. Validation of nomogram for diseasefree survival for colon cancer in UK population: a prospective cohort study.Int J Surg 2016;27:58–65.

29. Adam R, Delvart V, Pascal G, Valeanu A, Castaing D, Azoulay D, et al.Rescue surgery for unresectable colorectal liver metastases downstaged bychemotherapy: a model to predict long-term survival. Ann Surg 2004;240:644–57.

30. Andr�e T, Boni C, Mounedji-Boudiaf L, Navarro M, Tabernero J, Hickish T,et al. Oxaliplatin, fluorouracil, and leucovorin as adjuvant treatment forcolon cancer. N Engl J Med 2004;350:2343–51.

31. Des Guetz G, Schischmanoff O, Nicolas P, Perret GY, Morere JF, Uzzan B.Does microsatellite instability predict the efficacy of adjuvant chemother-apy in colorectal cancer? A systematic review with meta-analysis. Eur JCancer 2009;45:1890–6.

32. Vilar E, Gruber SB. Microsatellite instability in colorectal cancer-the stableevidence. Nat Rev Clin Oncol 2010;7:153–62.

33. Jover R, Nguyen TP, Perez-Carbonell L, Zapater P, Paya A, Alenda C, et al.5-Fluorouracil adjuvant chemotherapy does not increase survival inpatients with CpG island methylator phenotype colorectal cancer.Gastroenterology 2011;140:1174–81.

34. Gunderson LL, Jessup JM, Sargent DJ, Greene FL, Stewart A. Revised tumorand node categorization for rectal cancer based on surveillance, epidemi-ology, and end results and rectal pooled analysis outcomes. J Clin Oncol2010;28:256–63.

35. NitscheU,MaakM, Schuster T, Kunzli B, Langer R, Slotta-Huspenina J, et al.Prediction of prognosis is not improved by the seventh and latest edition ofthe TNM classification for colorectal cancer in a single-center collective.Ann Surg 2011;254:793–800.

36. Zhang JX, Song W, Chen ZH, Wei JH, Liao YJ, Lei J, et al. Prognostic andpredictive value of a microRNA signature in stage II colon cancer: amicroRNA expression analysis. Lancet Oncol 2013;14:1295–306.

37. de Sousa EMF, Colak S, Buikhuisen J, Koster J, Cameron K, de Jong JH,et al. Methylation of cancer-stem-cell-associated Wnt target genes pre-dicts poor prognosis in colorectal cancer patients. Cell Stem Cell2011;9:476–85.

38. Arpino G, Generali D, Sapino A, Del Mastro L, Frassoldati A, de LaurentisM, et al. Gene expression profiling in breast cancer: a clinical perspective.Breast 2013;22:109–20.

39. Marusyk A, Almendro V, Polyak K. Intra-tumour heterogeneity: a lookingglass for cancer? Nat Rev Cancer 2012;12:323–34.

40. Zhang ZY, LuoQF, YinXW,Dai ZL, Basnet S, GeHY.Nomograms to predictsurvival after colorectal cancer resection without preoperative therapy.BMC Cancer 2016;16:658.

41. Kawai K, Sunami E, Yamaguchi H, Ishihara S, Kazama S, Nozawa H,et al. Nomograms for colorectal cancer: a systematic review. WorldJ Gastroenterol 2015;21:11877–86.

42. Abdul Aziz NA, Mokhtar NM, Harun R, Mollah MM, Mohamed Rose I,Sagap I, et al. A 19-gene expression signature as a predictor of survival incolorectal cancer. BMC Med Genomics 2016;9:58.

43. Shi M, He J. ColoFinder: a prognostic 9-gene signature improvesprognosis for 871 stage II and III colorectal cancer patients. PeerJ2016;4:e1804.

44. Xu G, Zhang M, Zhu H, Xu J. A 15-gene signature for prediction of coloncancer recurrence and prognosis based on SVM. Gene 2017;604:33–40.

45. Nguyen MN, Choi TG, Nguyen DT, Kim JH, Jo YH, Shahid M, et al.CRC-113 gene expression signature for predicting prognosis in patientswith colorectal cancer. Oncotarget 2015;6:31674–92.

46. Wang L, Shen X, Wang Z, Xiao X, Wei P, Wang Q, et al. A molecularsignature for the prediction of recurrence in colorectal cancer. Mol Cancer2015;14:22.

47. Vargas T, Moreno-Rubio J, Herranz J, Cejas P, Molina S, Gonzalez-VallinasM, et al. ColoLipidGene: signature of lipid metabolism-related genes topredict prognosis in stage-II colon cancer patients. Oncotarget 2015;6:7348–63.

48. Sveen A, Agesen TH, Nesbakken A, Meling GI, Rognum TO, Liestol K, et al.ColoGuidePro: a prognostic 7-gene expression signature for stage IIIcolorectal cancer patients. Clin Cancer Res 2012;18:6001–10.

49. Salazar R, Roepman P, Capella G, Moreno V, Simon I, Dreezen C, et al.Gene expression signature to improve prognosis prediction of stage II andIII colorectal cancer. J Clin Oncol 2011;29:17–24.

50. KaladyMF, Dejulius K, Church JM, Lavery IC, Fazio VW, Ishwaran H. Genesignature is associated with early stage rectal cancer recurrence. J Am CollSurg 2010;211:187–95.

51. Peng J, Wang Z, Chen W, Ding Y, Wang H, Huang H, et al. Integration ofgenetic signature and TNM staging system for predicting the relapse oflocally advanced colorectal cancer. Int J Colorectal Dis 2010;25:1277–85.

52. Watanabe T, Kobunai T, Yamamoto Y, Kanazawa T, Konishi T, Tanaka T,et al. Prediction of liver metastasis after colorectal cancer using reverse






transcription-polymerase chain reaction analysis of 10 genes. Eur J Cancer2010;46:2119–26.

53. Hao JM, Chen JZ, Sui HM, Si-Ma XQ, Li GQ, Liu C, et al. A five-genesignature as a potential predictor of metastasis and survival in colorectalcancer. J Pathol 2010;220:475–89.

54. Watanabe T, Kobunai T, Sakamoto E, Yamamoto Y, Konishi T, Horiuchi A,et al. Gene expression signature for recurrence in stage III colorectal cancers.Cancer 2009;115:283–92.

55. Garman KS, Acharya CR, Edelman E, Grade M, Gaedcke J, Sud S, et al. Agenomic approach to colon cancer risk stratification yields biologicinsights into therapeutic opportunities. Proc Natl Acad Sci U S A2008;105:19432–7.

56. Lin YH, Friederichs J, Black MA, Mages J, Rosenberg R, Guilford PJ, et al.Multiple gene expression classifiers from different array platformspredict poor prognosis of colorectal cancer. Clin Cancer Res 2007;13:498–507.


Xiong et al.




2018;16:1373-1384. Published OnlineFirst May 21, 2018.Mol Cancer Res Yongfu Xiong, Wenxian You, Min Hou, et al. Improves Prognosis Prediction for Colorectal CancerNomogram Integrating Genomics with Clinicopathologic Features

Updated version

10.1158/1541-7786.MCR-18-0063doi:

Access the most recent version of this article at:

Material

Supplementary

http://mcr.aacrjournals.org/content/suppl/2018/05/19/1541-7786.MCR-18-0063.DC1

Access the most recent supplemental material at:

Cited articles

http://mcr.aacrjournals.org/content/16/9/1373.full#ref-list-1

This article cites 56 articles, 10 of which you can access for free at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://mcr.aacrjournals.org/content/16/9/1373To request permission to re-use all or part of this article, use this link



http://mcr.aacrjournals.org/lookup/doi/10.1158/1541-7786.MCR-18-0063http://mcr.aacrjournals.org/content/suppl/2018/05/19/1541-7786.MCR-18-0063.DC1http://mcr.aacrjournals.org/content/16/9/1373.full#ref-list-1http://mcr.aacrjournals.org/cgi/alertsmailto:[email protected]://mcr.aacrjournals.org/content/16/9/1373http://mcr.aacrjournals.org/

/ColorImageDict > /JPEG2000ColorACSImageDict > /JPEG2000ColorImageDict > /AntiAliasGrayImages false /CropGrayImages false /GrayImageMinResolution 200 /GrayImageMinResolutionPolicy /Warning /DownsampleGrayImages true /GrayImageDownsampleType /Bicubic /GrayImageResolution 300 /GrayImageDepth -1 /GrayImageMinDownsampleDepth 2 /GrayImageDownsampleThreshold 1.50000 /EncodeGrayImages true /GrayImageFilter /DCTEncode /AutoFilterGrayImages true /GrayImageAutoFilterStrategy /JPEG /GrayACSImageDict > /GrayImageDict > /JPEG2000GrayACSImageDict > /JPEG2000GrayImageDict > /AntiAliasMonoImages false /CropMonoImages false /MonoImageMinResolution 600 /MonoImageMinResolutionPolicy /Warning /DownsampleMonoImages true /MonoImageDownsampleType /Bicubic /MonoImageResolution 900 /MonoImageDepth -1 /MonoImageDownsampleThreshold 1.50000 /EncodeMonoImages true /MonoImageFilter /CCITTFaxEncode /MonoImageDict > /AllowPSXObjects false /CheckCompliance [ /None ] /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false /PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true /PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXOutputIntentProfile (None) /PDFXOutputConditionIdentifier () /PDFXOutputCondition () /PDFXRegistryName () /PDFXTrapped /False

/CreateJDFFile false /Description > /Namespace [ (Adobe) (Common) (1.0) ] /OtherNamespaces [ > /FormElements false /GenerateStructure false /IncludeBookmarks false /IncludeHyperlinks false /IncludeInteractive false /IncludeLayers false /IncludeProfiles false /MarksOffset 18 /MarksWeight 0.250000 /MultimediaHandling /UseObjectSettings /Namespace [ (Adobe) (CreativeSuite) (2.0) ] /PDFXOutputIntentProfileSelector /NA /PageMarksFile /RomanDefault /PreserveEditing true /UntaggedCMYKHandling /LeaveUntagged /UntaggedRGBHandling /LeaveUntagged /UseDocumentBleed false >> > ]>> setdistillerparams> setpagedevice

nomogram integrating genomics with clinicopathologic ..."colorectal cancer," "colon cancer,"...

Documents