决策树是通过递回分割

1、A Five-Gene Signature and Clinical Outcomein NonSmall-Cell Lung CancerFrom: n engl j med 356;1 january 4, 2007By: Hsuan-Yu Chen, M.Sc., Sung-Liang Yu, Ph.D., et alReporter: R6 謝廣宇 1BackgroundCurrent staging methods are inadequate for predicting outcomedeveloped a 5 -gene signaturenon small-cell lung

2、 cancer (NSCLC) most common cause of death from cancer worldwiderelapse rate with early-stage NSCLC is 40% within 5 years after potentially curative treatment23決策樹是通過遞迴分割（recursive partitioning）建立而成，遞迴分割是一種把資料分割成不同小的部分的疊代過程4cDNA microarray schemacDNA晶片製造原理5Microarray analysisOperation Principle:Samp

3、les are tagged with flourescentmaterial to show pattern of sample-probe interaction (hybridization)Microarray may have 60K probe6Microarray Processing sequenceFrom: Shin-Mu Tseng 7Gene-expression profiling by means of microarrays and reverse-transcriptase polymerase chain reaction (RT-PCR)examined g

4、ene expression in 125 surgical specimens of NSCLC, using microarrays and real-time RT-PCR to identify a gene signature correlated with clinical outcome8Methodscomputer-generated random numbers to assign specimens from 185 consecutive patients for microarray analysisfrozen specimens of lung-cancer ti

5、ssue from 125 randomly selected patients who underwent surgical resection of NSCLC at Taichung Veterans General Hospital between December 1999 and December 2003-five-gene risk-prediction model using an independent cohort of 60 randomly selected patients9125 specimens- 60 were adenocarcinomas, 52 wer

6、e squamous-cell carcinomas, and 13 were other types of cancernot received adjuvant chemotherapy672 genes associated with invasive activity, identified in a previous study- rearrayed in duplicate on a nylon membrane4 g of total RNA from each specimen10validate levels of expression of genes found on m

7、icroarray analysis RT-PCR was performed on 16 genes and a control gene for TATA-boxbinding protein (TBP), with use of specific TaqMan probes and primer sets - transcripts were amplified with reagent (TaqMan One-Step RT-PCR Master Mix Reagent, Applied Biosystems) and a sequence detection system (ABI

8、Prism 7900HT, Applied Biosystems)1112To reduce background noise background intensity values of less than 3000 were assigned value of 3000log-transformed to a base-2 scaleGenes with coefficients of variation of less than 3% were excluded from further analysesthe gene-expression intensity values were

9、transformed to ordinal coding values by ranking of level of gene expression among the 485 genes in 125 patients (60,625 observations)13intensity value was coded as 1-2-3-4 according to 0-25%-50%-75%-100%Hazard ratios from univariate Cox regression analysis to determine which genes associated with de

10、ath from any cause or recurrence of cancerProtective genes were defined with a hazard ratio for death of 114univariate Cox proportional-hazards regression analysisGenes significantly correlated with survival - a linear combination of gene-expression coding values weighted by regression coefficients

11、to calculate a risk score for each patientpatients risk score was calculated as sum of levels of expression of each gene, as measured by microarray analysis, multiplied by corresponding regression coefficients15high-risk gene signature or a low-risk gene signature, with 50th percentile (median) of t

12、he risk score as the threshold value (median, 4.9; range, 1.3 to 21.9)-threshold value to reflect almost half of patients with early stage NSCLC relapse within 5 years after potentially curative surgery, eliminate effect of extreme valueslevels of expression of 16 genes confirmed by RT-PCR and index

13、ed by Spearmans rank-correlation test further identified 5 genes significantly associated with survival16recursive-partitioning decision tree & Avadis software (Strand Genomic) a high-risk gene signature or a low-risk gene signatureKaplanMeier method to estimate overall survival and relapse-free sur

14、vivalMultivariate Cox proportional hazards regression analysis with stepwise selection evaluate independent prognostic factors with survival, 5-gene signature, age, sex, tumor stage, and histologic characteristics as covariates17To further validate our model, we applied it to microarray data from 86

15、 patients with NSCLC, reported by Beer et alfive genes (and their corresponding Affymetrix probe sets) were DUSP6 (X93920_at), MMD (X85750_at), STAT1 (M97936_at), ERBB3 (S61953_at), and LCK (M26692_s_at); the control gene was TBP (X54993_s_at)maximum follow-up time for survival analysis in our study

16、 was 62 months, we used 5-year survival data for 86 patients18Results16 genes correlated with death from any cause: 4 were protective genes (hazard ratio 1)1920dual-specificity phosphatase 6 (DUSP6) monocyte-to-macrophage differentiation associated protein (MMD)signal transducer and activator of tra

17、nscription 1 (STAT1) v-erb-b2 avian erythroblastic leukemia viral oncogene homolog 3 (ERBB3) lymphocyte-specific protein tyrosine kinase (LCK)21225-gene signature strongly associated with overall survival (sensitivity, 98%;specificity, 93%; positive predictive value, 95%; negative predictive value,

18、98%; and overall accuracy, 96%)Among patients with stage II disease, overall survival did not differ significantly between high-risk and low-risk gene signatures, probably owing to small number of patients232425DiscussionNSCLC is a heterogeneous disease based on clinical and pathological findings ma

19、y have limit of usefulness for predicting outcomes, but molecular methods add valueuse of microarrays in clinical practice limited by the large number of genes in the analysis ,complicated methods, lack of reproducibility and independent validation of results, and need for fresh-frozen tissue26RT-PC

20、R involving a small number of genes allows for accurate and reproducible quantification of results for RNA obtained from small amounts of paraffin- embedded specimens125 frozen tumor specimens from patients with NSCLC randomly divided into a training set (63specimens) and a testing set (62 specimens

21、)27selection of genes in microarray training set was validated in microarray testing set, and patterns of gene expression found on microarray analysis were validated by RT-PCRresults in our Chinese patients were also validated with use of a set of published NSCLC microarray data from patients from a

22、 Western population with NSCLC28propose who have tumors with a high-risk gene signature could benefit from Cisplatin-based adjuvant chemotherapy, those with a low-risk gene signature could be spared - large-scale, multicenter studies are necessary to test this idea29STAT1 arrested growth and apoptosis in many types of cancer cells by inducing expression of p21WAF1 and caspase.MMD expressed in mature macrophages, macrophage activation promotes cancer metastasis, BUT function of MMD protein is unknownDUSP6 inactivates extracellular signal-regulated kinase 2 (ERK2) (also known a