6. Thoughts on prognostic analysis of iron death factor (3+)

Hello everyone, what I share with you today is the recent publication in International Journal of Biological Sciences (IF: 4.
858).
The article mainly talks about constructing the prognostic signature of cancer based on iron death-related genes.

The classic way of constructing prognostic risk signatures in this way has been received and published, and the public accounts have been pushed countless.
The idea is clear, simple and complete.
It is nothing more than: 1.
Find a gene set of interest; 2.
Dimensionality reduction to screen prognostic factors; 3.
Dimensionality reduction again to construct the prognostic risk signature; 4.
Effectiveness verification Barabara, etc.

Not much to say, come and read it carefully.

Shengxin people provide professional and reproducible iron death information analysis.
Interested in scanning code, pre-determined ideas, iron death-related gene signature construction to predict overall liver cancer survival 1.
Abstract Liver cancer is a highly heterogeneous cancer, and iron death (Ferroptosis) is a type Iron-dependent form of cell death (iron death can be induced by sorafenib).

The prognostic value of iron death-related genes in liver cancer remains to be further studied.

Based on the TCGA data set, this study screened the prognostic-related difference factors, and constructed a 10-gene cancer prognostic signature through the Lasso-cox regression model, which divided liver cancer into high- and low-risk groups.
The OS of the high-risk group was significantly lower than that of the low-risk group ( p<0.
001).

In addition, the study used the ICGC liver cancer data set for prognostic signature verification, and similar results were also obtained (p=0.
001).

ROC curve analysis also verified the prognostic efficacy of the signature.

Flow chart: Figure 1: Flow chart two, data and method data set analysis Data set: TCGA-LIHC 371 liver cancer patients with RNA-seq data and clinical information (https://portal.
gdc.
cancer.
gov/repository); verification Data set: ICGC (LIRI-JP) 231 RNA-seq data and clinical information of liver cancer samples (https://dcc.
icgc.
org/projects/LIRI-JP); see Table 1 for clinical information of TCGA and ICGC data sets.

Table 1: The construction and verification of the clinical information table of the training set and the validation set of the prognostic iron death-related gene signature.
First, screen the differential gene DEGs between TCGA liver cancer and paracarcinoma (R package: limma, p-adj<0.
05); secondly, use Univariate cox regression analysis was used to screen iron death factors related to the prognosis of liver cancer (P value corrected by Benjamini & Hochberg (BH)).

Then take the intersection with DEGs to obtain candidate genes for constructing signature, use String database to construct PPI protein interaction network based on these candidate genes, and study the interaction of candidate factors; finally, use lasso regression analysis to further screen the prognostic factors of liver cancer (R package: glmnet ), calculate the patient risk score according to the expression of each candidate factor and lasso regression coefficient: score = esum (each gene's expression × corresponding coefficient).

According to the median score, liver cancer samples are divided into high-risk and low-risk groups.
R package: stats is used for principal component analysis (PCA), and R package: Rtsnet-SNE is used for grouping visualization.

The survival analysis of each candidate gene expresses the best cut point by the R package: survminer, and the ROC curve analysis uses the R package: survivalROC to evaluate the predictive ability of the signature.

 Functional enrichment analysis The GO function and KEGG pathway enrichment analysis of the differentially expressed genes DEGs in the high and low risk groups (R package: clusterProfiler); 16 immune cell infiltration scores and 13 immune-related pathway activity scores are enriched by a single sample gene set Analysis (ssGSEA, R package: gsva) is performed.

 Statistical analysis All statistical analysis is based on R software (Version 3.
5.
3) or SPSS (Version 23.
0).
3.
Result analysis 1.
TCGA identifies iron death-related differential genes.
81.
7% of iron death-related genes show differences between cancer and normal.
Among them, 27 are related to the prognosis of liver cancer (Figure 2a).

It should be noted that HMOX1 is significantly up-regulated in cancer, but it shows a good prognosis.
High expression of other genes is associated with a worse prognosis.
Therefore, this gene is deleted and the remaining 26 iron death factors are used for subsequent analysis (Figure 2b- c).

The PPI interaction network was constructed based on these candidate factors, with GPX4, G6PD and NQO1 as the core genes (Figure 2d), and correlation analysis of iron death prognostic factors (Figure 2e).

Figure 2.
TCGA to identify candidate iron death-related genes.
2.
TCGA data set to construct a prognostic model.
Based on the Lasso and cox regression analysis, based on the 26 iron death candidate factors of TCGA, a prognostic signature containing 10 genes was finally constructed (the formula is more complicated here.
Don't let it go, if you want to read it, you can contact customer service for literature), and divide the sample into high-risk (182) and low-risk (183) groups according to the median (Figure 3a).

The distribution of high and low risk samples in clinical indicators is shown in Table 2.

The PCA and t-SNE analysis results show that the signature can distinguish the two groups of samples well (Figure 3b-c).
The risk of death in high-risk patients is significantly higher than that in low-risk patients, and OS survival is worse (Figure 3e, p<0.
001) ).

The ROC curve assesses the OS predictive efficacy AUC value of the risk score (0.
8 in 1 year, 0.
69 in 2 years, 0.
668 in 3 years). Figure 3.
Prognostic analysis of the TCGA 10-genes signature model 3.
Prognostic signature ICGC data set verification Based on the ICGA data set, the survival analysis of the 10 genes contained in the signature is related to worse OS except CARS.

Patients in the ICGC cohort were also divided into high- and low-risk groups according to the median calculated by the same formula as the TCGA cohort to test the robustness of the model constructed by the TCGA cohort.

Similar results to TCGA were obtained.
The distribution of ICGC high and low risk samples in clinical indicators is shown in Table 2.

The PCA and t-SNE analysis results also show that the signature can distinguish the two groups of samples well (Figure 4b-c), and high-risk patients with lower-risk patients have a shorter survival time (Figure 4e, p=0.
001).

The ROC curve assesses the OS predictive efficacy AUC value of the risk score (0.
68, 0.
69, 0.
718 for 1, 2, and 3 years, respectively).

 Figure 4.
ICGC 10-genes signature model verification Table 2: TCGA and ICGC cohort high- and low-risk group clinical index characteristics 4.
10-genes signature model independent prognostic value single factor, multivariate cox regression analysis to determine the risk score is the independent prognosis of liver cancer OS index.

Univariate cox regression analysis found that the risk score in the training and validation data sets were significantly correlated with OS prognostic factors for liver cancer (Figure 5a-b, TCGA p<0.
001, ICGC p=0.
006).

After adjusting for other confounding factors, in the multivariate Cox regression analysis, risk score was still proved to be an independent predictor of OS (Figure 5a-b, TCGA p<0.
001, ICGC p=0.
005).

Figure 5.
TCGA and ICGC data set risk scores are independent prognostic factors for liver cancer.
V.
Functional analysis of TCGA and ICGC data sets.
In order to clarify the biological functions and pathways related to risk scores, GO and KEGG enrichment were performed on the differential genes of the high and low risk groups.
Set analysis. As expected, DEGs were enriched in several biological processes related to iron (Figure 6a.
c).

Interestingly, the DEGs of TCGA were also significantly enriched in many immune-related biological processes (Figure 6a), and similar results were obtained in the ICGC cohort (Figure 6c).

KEGG pathway analysis showed that the cytokine-cytokine interaction pathway was enriched in both data sets (Figure 6bd).

 Figure 6.
GO and KEGG enrichment analysis results TCGA (ab), ICGC (cd) 6.
Analysis of the correlation between risk score and immune infiltrating cells In order to further study the correlation between risk score and immune status, the study used ssGSEA to quantify the scores of immune cells and Immune-related functions and pathway scores.

The results showed that the scores related to the antigen presentation process in the TCGA cohort were significantly different between the high and low wind groups (P-adj <0.
05, Figure 7a-b), and the high risk had a higher score for the cytokine-cytokine interaction pathway (Figure 7b).

In the high-risk group, type II IFN response, type I IFN response, and NK cell scores were lower, and immune check site activity, while macrophages and Treg scores were the opposite (P-adj <0.
05, Figure 7a-b).

The ICGC cohort verified the differences in HLA, MHC type I, IFN type II response, checkpoint molecules, macrophages, and Treg cells between the two risk groups (corrected P<0.
05, Figure 7c-d).

In particular, the macrophage scores in the two data sets have the largest statistical difference, which is consistent with the findings in the GO analysis.

 Figure 5.
TCGA(ab)he ICGA(cd) high and low risk group ssGSEA differences (ac-16 immune cell scores, cd-13 immune function scores) Summary: This is the end of today’s sharing, there is really no picture Many, do you have any friends you don’t understand. If you don’t understand it, you should reflect on it.
After all, such a prognostic story public account has been repeatedly shared many times! Let me repeat it again: To be punctual (to study the gene set specifically related to cancer type), to use the data well (analysis + validation samples should not be too small, the results are better), it's almost the same.

If you are interested in sharing today, would you like to tease? Shengxin people provide professional, reproducible iron death life letter analysis and interested scan code reservation ideas