Please note that we have stopped the regular imports of Gene Expression Omnibus (GEO) data into ArrayExpress. This may not be the latest version of this experiment.
E-GEOD-8894 - Transcription profiling of human NSCLC samples to predict recurrence-free survival in postoperative nsclc patients
Released on 17 June 2008, last updated on 10 June 2011
Background:; One of the main fields of lung cancer research is identifying patients who are at high risk of post-resection recurrence. Individual recurrence risk evaluation by accurate but simple and reproducible method is needed for the clinical practice. Results:; The log-rank test and further selection by our criteria of assayability generated 87 genes from microarray data with significant level 5%. Of these, by PTQ-PCR, the expression of most significant 18 genes was obtained. Using these gene expression information and clinical parameters, by stepwise variable selection method, the recurrence prediction model, which composed of 6 genes (CALB1, MMP7, SLC1A7, GSTA1, CCL19, IFI44) and pStage and cell differentiation, were developed. Validation into the two independent cohorts showed good results of the proposed model (p=0.0314, 0.0305, respectively). The predicted median recurrence-free survival times for each patient were reflected real ones well. Conclusions:; Our method of individualized recurrence risk prediction is accurate, technically simple and reproducible to be used in clinical practice. Therefore, it would be useful in customizing the lung cancer management strategies. Experiment Overall Design: Methods: Experiment Overall Design: At first, we selected the statistically significant genes from the analysis of time-to-recurrence and censoring information from 138 whole-genome wide microarray data. Then, we further reduced the number of genes which could be reliably reproducible by RTQ-PCR. With these assayable genes and clinical parameters, construction of recurrence prediction model by Cox proportional hazard regression was done. After validation into two independent cohorts (n=59 and n=56), the model was transformed into recurrence prediction for the each patient.
transcription profiling by array, unknown experiment type