Supplementary MaterialsData_Sheet_1. et al., 2018) and DCDB (Liu Y. et al., 2014) database. The targets of anti-cancer drugs were extracted from drugBank (Wishart et al., 2018) and TTD (Li et al., 2018) database. Abstract Background The dysregulation of non-coding RNAs (ncRNAs) such as miRNAs and lncRNAs are associated with the pathogenesis and progression in multiple cancers including solid tumors. Comprehensive investigations of prognosis-related ncRNA markers could promote the development of therapeutic strategies for solid tumors, but rarely reported. Methods By taking advantage of The Cancer Genome Atlas (TCGA), pan-cancer prognosis analysis (PCPA) models were firstly constructed based on miRNA and lncRNA expression profiles of 8,450 samples in 19 solid tumors. Further, the co-occurrence and exclusivity among ncRNA markers were systematically analyzed for different cancers. Results In identified ncRNA makers, 71% of the miRNA markers were shared in multiple cancers, whereas 96% of the lncRNA markers were cancer-specific. Moreover, to analyze the regulation patterns of prognosis-related ncRNAs at the pan-cancer level, miRNA Betanin ic50 markers were further annotated into eight carcinogenic pathways. Results represented that approximately 86% of these miRNA markers could regulate the PI3K-Akt signaling pathway, while only 48% Rabbit polyclonal to LRIG2 for the Notch signaling pathway. Finally, among 126 common genes that participated in eight carcinogenic pathways, BCL2, CSNK2A1, EGFR, PDGFRA, and VEGFA were proposed as potential drug targets for multiple cancers. Conclusion The prognosis analysis and regulation characteristics of ncRNAs presented in this study may help to facilitate the discovery of anti-cancer drugs for multiple solid tumors. 0.01 and absolute fold change value | FC| 2, and the DE miRNAs were filtered with 0.05. The DE ncRNAs in each cancer type was the combination of the DE miRNAs and lncRNAs, and the samples were the intersection of samples in miRNAs and lncRNAs expression profiles of the corresponding cancer. Note that, the number of training samples is quite small in merged ncRNAs for KICH (45 samples) and LUSC (44 samples), which may lead to the overfitting of PCPA modeling. Thus, factor analysis was performed to reduce the dimensions of the combined ncRNAs in KICH and LUSC by the package 1.9.12.31 of R software (Lorenzo-Seva and Van Ginkel, 2016b), and the top 10 lncRNAs for each factor were selected according to the weight matrix to identify the prognosis-related ncRNAs. The expression profiles of DE miRNAs, lncRNAs, as well as combined ncRNAs of each cancer, were used for subsequent modeling. Construction of Pan-Cancer Prognosis Evaluation (PCPA) Model Teaching and tests datasets of every solid tumor had been acquired through the spatial subset sampling solution to generate PCPA versions. Typically, the 1st test A was chosen as the seed, and the next sample B using the farthest spatial range from test A was chosen. Next, Betanin ic50 the 3rd sample using the farthest typical range toward both examples A and B was extracted. After that, sampling was repeated until two-thirds from the negative and positive examples had been screened as working out set, and the others examples had been thought as the testing set. Both single-omic and two-omic ncRNA datasets of 19 solid tumors were used to construct the PCPA model. Here, four machine learning models including NN, NB, LR, and SVMs were implemented by using the python 2.7.9 package 0.3.6 (Lorenzo-Seva and Van Ginkel, 2016a) to generate the PCPA model based on labels divided from the median OS of corresponding patient samples. Survival Analysis Survival analysis (Wang et al., 2019) was performed based on the classification results of different PCPA models. KM survival curves of different Betanin ic50 samples were evaluated by using the R 3.1-11 and 0.4.6 package (Modhukur et al., 2018). In addition, the log-rank test (Rantala et al., 2019) was employed to test the difference between the two compared sample groups. Construction of Refined Gene-Specific Pathway Genes regulated by corresponding prognosis-related miRNA markers were obtained from miRNA-target interaction databases, including miRTarbase 7.0 (Chou et al., 2018), miRecords 2013 (Xiao et al., 2009), and TargetScan 3.1 (Riffo-Campos et al., 2016). Genes that were regulated by prognosis-related lncRNA markers were converted Betanin ic50 from gene ENSEMBL to gene SYMBOL by Betanin ic50 the 3.7.0 (Prummer, 2019) and the 3.14.3 (Yu et al., 2012) package in R software. Further, eight canonical signaling pathways with frequent genetic alterations in cancers regulated (Sanchez-Vega et al., 2018) by the above detected prognosis-related.
Categories