Page 31 - GPD-2-3
P. 31

Gene & Protein in Disease                                           Signatures construction strategies for TC
















































            Figure 5. The flow chart of strategy 3: Signatures based on methylation-related molecules. CoxPH model* means the univariate and multivariate Cox
            proportional hazards regression analysis. DEmethylation, differentially expressed methylation.
            divided  into  the  BRAFWT  group  and  the  BRAFV600E   data, we divided the main validation methods into two
            group according to the mutation information to identify   categories: internal validation and external validation.
            DEGs. Furthermore, weighted correlation network      Data used to construct a signature were classified into
            analysis (WGCNA) function or pathway enrichment, and   three main types: the whole dataset, the training set, and
            correlation analysis was used to construct a co-expression   the testing set. According to the research purpose, the
            network. In general, 8 BRAFV600E-associated prognostic   verification methods of signatures include the following
            biomarkers (FN1, MET, SLC34A2, NGEF, TBC1D2,       three categories: (1) the validation methods of prognostic
            PLCD3, PROS1, and NECTIN4) were discovered, which   signatures were performed in the whole dataset without
            could be considered as targets for TC treatment .  distinguishing training and testing set [34,42,47,78] ; (2) the
                                                  [53]
                                                               process of validation was conducted in the training set
            3.4. The verification methods of signatures        and testing set, respectively [26,28,41] ; and (3) distinguish
            In this review, the methods of constructing prognostic   among the entire dataset, training set, and testing set, and
            signatures were summarized. It is noteworthy that the   validate  signatures  in  all  three  datasets [44,66] .  In  terms  of
            validation effects of the constructed signatures could   specific verification methods, most researchers adopted
            reflect  the  potential  of  prognostic  signatures  for  clinical   common aforementioned validation approaches, including
            application. Through retrospective investigation and   Kaplan-Meier  survival,  ROC curve,  and  univariate and
            integrated analysis, we found that the verification methods   multivariate Cox regression analyses. The main methods
            described in this review were often closely related to the   of internal validation were summarized above.
            application of data acquisition and the process of signature   A common approach to conducting external validation
            construction.  Therefore, according to the  application of   was to employ hospital data sources as the validation


            Volume 2 Issue 3 (2023)                         15                       https://doi.org/10.36922/gpd.1138
   26   27   28   29   30   31   32   33   34   35   36