Discussion
To sum up, this article validated the prediction role of CAF infiltration in the prognosis of OS, which might shed light on the treatment of OS.
Methods
The processed RNA-seq data and the corresponding clinical and molecular information were retrieved from the Cancer Genome Atlas Program (TCGA) database and processed data of tumor tissue was obtained from Gene Expression Omnibus (GEO) database. Xcell method was used in data processing, and Gene set variation analysis (GSVA) was used to calculates enrichment scores. Nomogram was constructed to evaluate prognostic power of the predictive model. And the construction of risk scores and assessment of prognostic predictive were based on the LASSO model.
Results
This study classified Cancer Genome Atlas (TCGA) cohort into high and low CAFs infiltrate phenotype with different CAFs infiltration enrichment scores. Then TOP 9 genes were screened as prognostic signatures among 2,488 differentially expressed genes between the two groups. Key prognostic molecules were CGREF1, CORT and RHBDL2 and the risk score formula is: Risk-score = CGREF1*0.004 + CORT*0.004 + RHBDL2*0.002. The signatures were validated to be independent prognostic factors to predict tumor prognosis with single-factor COX and multi-factor COX regression analyses and Norton chart. The risk score expression of risk score model genes could predict the drug resistance, and significant differences could be found between the high and low scoring groups for 17-AAG, AZD6244, PD-0325901 and Sorafenib.