Ferroptosis

To obtain literature on ferroptosis, we searched the PubMed database (https://www.ncbi.nlm.nih.gov/pubmed) using the term ‘ferroptosis’ on 12 July 2019. When our manuscript was under review, we also searched the PubMed database on 20 February 2020 to find all ferroptosis articles of year 2019. All ferroptosis-related articles found in PubMed were downloaded. We then read these articles to identify genes, small molecules and diseases related to ferroptosis. Afterwards, we performed annotation for regulators, markers and effects of ferroptosis on disease according to experimental evidence from the collected publications.

Annotation data sets in FerrDb belong to three categories (Table 1). Genes were annotated as drivers, suppressors and markers. Small molecules were annotated as inducers and inhibitors. Drivers, suppressors, inducers and inhibitors are regulators of ferroptosis: drivers and inducers positively regulate ferroptosis, while suppressors and inhibitors negatively regulate ferroptosis. Markers do not regulate ferroptosis, but they indicate the occurrence of ferroptosis. Ferroptosis affects the development of disease in two ways. Ferroptosis was then annotated to either aggravate or alleviate an illness.
To be annotated as a ferroptosis regulator, genes and small molecules must possess explicit evidence to prove their regulatory role in ferroptosis. This kind of evidence is generally represented by an author statement of the role of the regulator in an original article. Genes that only undergo abundance, modification or stability change or are merely a component of a functional signaling axis or interaction network were annotated as markers. To annotate ferroptosis’ effect on diseases, evidence based on a growth test in cell lines or animal models was required.
In comparison with revealing a small molecule’s role, confirming a gene’s function is more challenging. We therefore dedicated more effort to gene annotation. A confidence level was assigned to each annotation to indicate its reliability (Table 2). Experimental reproducibility is correlated with results consistency, so the number of experiments was used as a score of the accuracy of the regulatory role of annotated genes. Critical cases (e.g. article retraction, conflicting results) that may affect the annotation reliability were highlighted by a caution statement. Other noteworthy information (e.g. inconsistent gene symbols) that seems less likely to impair annotation quality was denoted with a remark.

The "limma" R package was used to identify the differentially expressed genes (DEGs) between tumor tissues and adjacent nontumorous tissues with a false discovery rate (FDR) < 0.05 in the TCGA cohort. Univariate Cox analysis of overall survival (OS) was performed to screen ferroptosis-related genes with prognostic values. P values were adjusted by Benjamini & Hochberg (BH) correction. An interaction network for the overlapping prognostic DEGs was generated by the STRING database (version 11.0) 19 (link). To minimize the risk of overfitting, the LASSO-penalized Cox regression analysis was applied to construct a prognostic model 20 , 21 (link). The LASSO algorithm was used for variable selection and shrinkage with the "glmnet" R package. The independent variable in the regression was the normalized expression matrix of candidate prognostic DEGs, and the response variables were overall survival and status of patients in the TCGA cohort. Penalty parameter (λ) for the model was determined by tenfold cross-validation following the minimum criteria (i.e. the value of λ corresponding to the lowest partial likelihood deviance). The risk scores of the patients were calculated according to the normalized expression level of each gene and its corresponding regression coefficients. The formula was established as follows: score= e^{sum (each gene's expression × corresponding coefficient)}. The patients were stratified into high-risk and low-risk groups based on the median value of the risk score. Based on the expression of genes in the signature, PCA was carried out with the "prcomp" function of the "stats" R package. Besides, t-SNE were performed to explore the distribution of different groups using the "Rtsne" R package. For the survival analysis of each gene, the optimal cut-off expression value was determined by the "surv_cutpoint" function of the "survminer" R package. The "survivalROC" R package was used to conduct time‐dependent ROC curve analyses to evaluate the predictive power of the gene signature.

RNA-seq data and clinical information of 1091 BRCA samples and 113 normal samples were accessed from the Cancer Genome Atlas (TCGA) database. RNA-seq data of 50 BRCA patients with survival information were collected from the ICGC database (https://dcc.icgc.org/) and utilized for risk signature validation, namely validation set. 117 BRCA samples from GSE88770 dataset were also used to verify the prognostic risk model. 259 ferroptosis-related genes (FRGs) were extracted from the FerrDb database. 1542 RNA-binding proteins (RBPs) were derived from a previous report (Wang et al., 2021 (link)).