Explore The Mechanism Of Prostate Cancer Recurrence By RNA Sequencing And Whole Exome Sequencing

Prostate cancer is the common cancer among men. As to localized prostate cancer, the standard therapy is radical prostatectomy or radiotherapy, which is sometimes supplemented with hormone therapy. After that, a large amount of patients still experience recurrence, local recurrence or metastasis, after long periods of remission. With further research on prostate cancer, more and more biomarkers are being identified which have significantly improved the prediction of recurrence, such as AR splice events, TMPRSS2:EGR gene fusion, long noncoding rna MALAT-1 and SChLAPl. Nonetheless, the question why some patients would experience recurrence still puzzles the scientists, though a few ideas about mechanisms of prostate cancer recurrence has been put forward, such as mesenchymal to epithelial transition, reactivation of androgen signaling, loss of metastasis suppressor genes.So here we analyze the transcriptome sequencing data and whole exome sequencing data and try to find the difference at the molecular level between recurrence cancer and non-recurrence cancer, then conclude the mechanisms of prostate cancer recurrence and find more and moreeffective recurrence predictors.In the first part, we reanalysed transcriptome sequencing data of prostate cancer samples(recurrence or non-recurrence) from GEO database in many ways. At last, we identified an alternative splicing events of MYC gene in recurrence group, and it would change the protein sequence significantly and even affect the structure and function of the protein. What’s more, the protein participated in many significant biological process related with prostate cancer. In addition, we found 10 differentially co-expressed gene groups. In these groups, the expression of genes were highly correlated in non-recurrence group and show little or no correlation in recurrence group.Using inspect.group, we also identified the significantly changed pairwise correlation coefficients in differentially co-expressed groups. Using DCGL, we also selecteda few important regulators which contributed to the change in the regulatory mechanism between recurrence and non-recurrence group.At last, we also identified some specific somatic mutations in exons of recurrence group and these mutations would affect a few pathway related with cancer, such as cell cycle, ECM-receptor interaction, Focal adhesion, Ubiquitin-mediated proteolysis and pathways in cancer.In the second part, we collected 30 recurrent samples and 44 non-recurrent samples and did whole exome sequencing. Then we analysed sequencing data in many ways, such as single nucleotide variations, copy number variants, significantly mutated genes and classification model. Finally, we found 24 and 31 specific recurrent exon SNVs in recurrence group and non-recurrence group repectively and verified them in a large amount of samples. We also selected the best 50 variants to build randomforest model and the cross-validation error rate was as low as 5%. With mutsig, we filtered 191 and 189 significantly mutated genes in recurrence group and non-recurrence group repectively and these genes were mainly distributed in MAP kinase kinase kinase activity, actin filament, negative regulation of retinoic acid receptor signaling pathway, response to hyperoxia, notch signaling pathway and chronic myeloid leukemia. Some of pathway were related with cancer which might suggest that these pathways were vulnerable in tumorigenesis and play an important role in prostate cancer. In addition, we also identified 21 specific copy number loss regions and 190 specific amplified regions and the regions were mainly distributed in 1p3, 1q3,2q3,3p2,3q2,8q2,12ql,12q2. And the genes concerned mainly participated in pathway in cancer and IFN alpha signaling pathway.In a word, we found many specific variants in recurrence group, such as snp, cnv and alternative splicing, and the genes which contained these variants participated in many important biological process, including tumorigenesis and development. And we build an effective classification model whose cross-validation error rate was very low. So far, the specific relation between the variants and prostate cancer recurrence was still unclear and the accuracy of the model was not so stable when it was applied to other samples. So it was still necessary to further study the specific relations by experiment and test the accuracy of the model with a large amount of samples.