Page 20 - GPD-2-3

Gene & Protein in Disease Signatures construction strategies for TC

Figure 2. Evaluation of 28 prognostic signatures through meta-analysis. In the Forest plot, Qiuying Li, et al. 2017 means, the data of the multivariate Cox
a
regression analysis were derived from the training set; Kun Wang, et al. 2020 means, HR (95% CI) was extracted from multivariate regression analysis of
b
a 6-gene biomarker; and Kun Wang, et al. 2020 means, HR (95% CI) was extracted from multivariate analysis of a 5-lncRNA biomarker.
c
to identify differentially expressed genes and lncRNAs in the association between signatures and other clinical
two independent cohorts . Moreover, Wu et al. applied parameters [33,41,44] . Teng et al. used Kaplan-Meier survival
[21]
the LIMMA package ; (3) signature construction. As analysis as well as univariate and multivariate regression
[19]
differentially expressed mRNAs/lncRNAs/miRNAs were analyses to assess the prognostic power of the gene
identified, a series of analyses were applied to construct signature and whether it was considered an independent
signatures, including lease absolute shrinkage and prognostic factor . The flow chart of strategy 1 is shown
[21]
selection operator (lasso) regression, Cox regression, in Figure 3 and the summarized relevant information of
and functional enrichment analysis. Ruiz et al. employed signatures based on strategy 1 is presented in Table 1. For
a machine learning model with linear discriminant instance, Wu et al. constructed a gene signature using the
analysis (LDA) to construct a 25-gene signature of great method of DEGs screening mentioned in strategy 1 .
[19]
predictive ability . Wu et al. applied lasso regression First, by comparing the differences in expression profiling
[20]
analysis to identify DEGs significantly associated with between 510 tumors and 58 normal samples, 295 DEGs
PFI . Li et al. performed univariate Cox regression including 137 downregulated and 158 upregulated genes
[19]
analysis to obtain lncRNAs related to the survival of TC were found by the LIMMA package. Second, the authors
patients ; (4) signatures validation. Most researchers applied univariate Cox regression analysis to identify 50
[41]
applied ROC curves to reflect the predictive capabilities DEGs associated with PFI. The lasso regression model
of signatures [19,20,42] . Moreover, univariate and multivariate was then applied to develop a 5-gene prognostic signature
logistic regression analyses were also performed to validate (FXYD6, PLP2, FABP4, LYVE1, and TGFBR3). Finally, it is

Volume 2 Issue 3 (2023) 4 https://doi.org/10.36922/gpd.1138