Raw data were extracted from the database of Web of Science Core Collection. The literature search was performed within 1 day on 19 December 2021. The search formula was set as follows: ferroptosis (Topic) and “tumor*” or “cancer*” or “carcino*” or “onco*” (All Fields) and 2012–2021 (Year Published) and Meeting Abstracts or Editorial Materials or Early Access or Book Chapters or Corrections or News Items or Letters or Proceedings Papers or Retractions or Retracted Publications (Exclude—Document Types). A total of 1836 works of literature were obtained and exported for full record and cited references in the format of plain text files. Using Endnote software, three duplications were found. After removing duplications, a total of 1833 unique records remained, including 1,381 articles and 452 reviews (Supplementary Table S1).

CiteSpace software (Drexel University, Philadelphia, PA, United States) is a freely available Java application, which was widely used for visualizing and analyzing trends and patterns in the scientific literature (Chen, 2006). It was designed by Dr. Chaomei Chen in 2004 (Chen, 2004) and last updated on 17 January 2021 (Version 5.8.R2) (http://cluster.cis.drexel.edu/∼cchen/citespace/). Web of Science was the primary source of input data for CiteSpace. In this study, CiteSpace was used to get the map of network visualization (Countries/Regions, Institutions, Authors), the timeline view of co-occurrence (Keywords), and citation bursts (References and Keywords). The primary parameters were set as follows: time slicing (2012–2021), years per slice (1 year), selection criteria (g-index, k = 25), and pruning (minimum spanning tree, pruning sliced networks). Other parameters were set according to the CiteSpace manual for different situations.

VOSviewer software is a useful tool for constructing and visualizing bibliometric networks (van Eck and Waltman, 2010). It was developed by the Center for Science and Technology Research at Leiden University (Netherlands) in 2007 (van Eck and Waltman, 2007). The latest version 1.6.17 was released on 22 July 2021 and is free for download (https://www.vosviewer.com/). Co-authorship networks, citation-based networks, and co-occurrence networks can be created based on data downloaded from the Web of Science. In this study, VOSviewer was used to get the density map of co-cited authors, co-cited journals, and the co-occurrence of keywords.

Other information such as impact factor (IF) and Journal Citation Reports (JCR) division of journals were obtained from the Web of Science website directly on 19 November 2021. Microsoft Office Excel 2016 was used to analyze the annual publications. GraphPad Prism version 8 was used for statistical analysis. Pearson’s correlation analysis was used to investigate relationships between the annual publication numbers and annual citation numbers. A p value of <0.05 was considered statistically significant.

From the first paper published on 25 May 2012 to 19 December 2021, a total of 1833 publications related to ferroptosis in cancer were identified. Figure 1; Table 1 showed the annual number of publications of ferroptosis in cancer. In general, the annual number of publications increased year by year. The early stage (2012–2017) saw a relatively stable growth, with a publication counts of no more than 100 per year. While a steep growth of annual publication numbers was observed since 2018, approximately doubling every year. Up to date (19 December 2021), the publication outputs have reached over 846 in 2021. Compared to other fields of ferroptosis, the publications on ferroptosis in cancer account for most of the total publications annually (Table 1; Supplementary Figure S1). The prediction analysis indicated that over 3,500 papers will be published in 2025 (Supplementary Figure S2). As the number of citations also reflects research progress in a field (Waltman, 2016), citation analysis was performed in parallel. Similar to the trend of annual publication numbers, the annual citation numbers increased exponentially in the past decade (Figure 1). Indeed, the annual number of citations and publications was significantly correlated (Pearson r = 0.9994, p-value < 0.0001) (Supplementary Figure S3).

VOSviewer co-authorship analysis among the countries/regions was conducted to generate a network visualization map. Geographically, a total of 1833 articles were published from 63 different countries/regions covering five continents. Notably, Taiwan and the People’s Republic of China were merged into China; England, Scotland, Northern Ireland, and Wales were merged into the United Kingdom. To make the network clear, only 23 selected countries/regions that published at least 10 documents were visualized in Figure 2A. A complete list of documents, citations, and total link strength of the 63 countries/regions was displayed in Supplementary Table S2. Each node represented a different country/region. The size of the nodes is determined by the number of publications (The higher the number, the larger the node). The same color represented the same cluster. The line between the nodes represented the co-authorship between countries/regions (The stronger the cooperation relationship, the wider the line). The number of total link strengths reflected the total co-authorship strength between countries/regions.

As we can see from these results, there were 14 countries/regions that published only one paper and four countries/regions that published over 100 papers. Among the top 10 most prolific countries/regions (Table 2), China had the highest number of publications (1,092, 59.57%), followed by United States (489, 26.68%), Germany (151, 8.24%), and Japan (102, 5.56%). The United States had the highest number of citations (40,675), total link strength (416), and Germany had the highest number of Centrality (0.71). While Australia obtained the highest number of citations/documents (100.36) Here, high centrality (≥0.10) indicates active cooperation. Most of countries/regions cooperated with others (n = 60, total link strength≥1). The United States had the highest link strength (416), followed by China (333), Germany (226), and France (140) (Table 2). Notably, the total link strength between China and United States reached 159 (data not shown), indicating that these two countries with top publication outputs collaborated closely.

A total of 1737 institutions were identified that contributed to the 1833 papers. The top 10 most prolific institutions are shown in Table 3. Also, a complete list of documents, citations, and total link strength numbers of the 1833 institutions was displayed in Supplementary Table S3. As we can see, all of the top 10 institutions came from China (n = 7) and the United States (n = 3). Central South University (92, 5.02%) is the leading institute ranked by the number of publications, followed by Guangzhou Medical University (72, 3.93%) and Columbia University (67, 3.66%). Columbia University contributed the most citations (16,996) and had the highest number of citations/document (253.67). Guangzhou Medical University had the highest number of total link strengths (287) and Zhejiang University showed the highest centrality (0.24).

The co-authorship map of the top 30 selected institutions that published at least 27 documents was visualized in Figure 2B. The map had a total of 30 terms, four clusters, 136 links, and a total link strength of 449. Each node represented a different institution. The size of the nodes was determined by the number of publications (The higher the number, the larger the node). The same color represented the same cluster. The line between the nodes represented the co-authorship between institutions (The stronger the cooperation relationship, the wider the line). The number of total link strengths reflected the total co-authorship strength between institutions. From the visualization map, we can see that there had active cooperation among these 30 institutions within and between clusters. The top three institutions with the largest total link strength among the 1833 institutions are listed as follows: Guangzhou Medical University (n = 287), University of Pittsburgh (n = 272), and Columbia University (n = 204) (Supplementary Table S3).

A total of 10,032 authors published literature on ferroptosis with 87 authors having papers over five(Supplementary Table S4). The author visualization map was created by Citespace (Figure 3) and the top 10 authors ranked by publication frequency and centrality were shown in Table 4. In the cooperative network map (Figure 3), the nodes represented authors (the larger circle, the higher the number of publications); lines between the nodes represented a collaboration between two authors on the same article (the wider lines, the more frequency of collaborations); the colors in the nodes represented the years, and the purple ring represented high centrality (≥0.10). As shown in these data, Daolin Tang published the largest number of papers (n = 64), followed by Rui Kang (n = 58) and Brent R Stockwell (n = 48). Both Brent R Stockwell and Scott J Dixon published earliest (in 2012) and had the highest centrality (0.20).

Co-citation was defined as the frequency with which two documents were cited together by another or more papers at the same time, providing a way to study the specialty structure of science. In this study, a total of 45,521 co-cited authors were identified with 99 co-cited over 100 times (Supplementary Table S5). The co-citation density map showed the authors (n = 46) with co-citations of at least 150 and the top 10 co-cited authors were presented in Table 4. Scott J Dixon was the most frequently co-cited author (2,134 times), followed by Wan Seok Yang (1722 times) and Brent R Stockwell (795 times). It was worth noting that Scott J Dixon and Brent R Stockwell were among the top 10 productive authors and co-cited authors.

A total of 515 academic journals had published the 1833 papers on ferroptosis in cancer. The top 10 journals ranked by the number of publications were shown in Table 5, and a complete list of the 515 journals was displayed in Supplementary Table S6. Among the top 10 journals, Biochemical and Biophysical Research Communications (n = 54) and Frontiers in Oncology (n = 54) published the largest number of papers, followed by Frontiers in Cell and Developmental Biology (n = 52), and Cell Death & Disease (n = 51). Nine (90%) of the journals had an IF 2020 over five and seven (70%) were located in JCR (2020) Q1 region. Biochemical and Biophysical Research Communications had the largest number of total citations (1902 times). As shown in Table 6, the top three journals with the largest co-citations were Cell (co-citation: 5,342; IF 2020: 41.582), Nature (co-citation: 4,725; IF 2020: 49.962), Proceedings of The National Academy of Sciences of The United States of America (co-citation: 3,040; IF 2020: 11.205). Despite having few publications (n = 7) and a low impact factor (IF 2020: 5.157), the papers published in the Journal of Biological Chemistry obtained a large number of co-citations (2,882 times).

A total of 20 most cited publications on ferroptosis in cancer were listed in Table 7. Of these 20 publications, fourteen were articles (70%) and six were reviews (30%). All the top 20 references were cited more than 400 times. The article (entitled Ferroptosis: An Iron-Dependent Form of Nonapoptotic Cell Death) published in 2012 by Dixon, Scott J et al. had the largest number of citations (3,009). The purpose of burst detection was to determine whether an entity had significantly increased its presence within a certain period(Kleinberg, 2003). Through CiteSpace analysis, the top 25 references with the strongest citation bursts were identified (Figure 4). The result showed the first burst of reference began in 2013, and the most recent reference with a citation burst was observed in 2017. Among these 25 references, six references (24%) had a duration until 2021. The reference with the strongest burstness (strength = 128.73, from 2014 to 2019) was published in Cell by Yang, Wan Seok et al., followed by the reference published in Cell by Dixon, Scott J et al. (strength = 88.88, from 2013 to 2017) and Nature Cell Biology by Angeli, Jose Pedro Friedmann et al. (strength = 84.29, from 2015 to 2019).

To investigate the research hotspots in the field of ferroptosis in cancer, all keywords (including author keywords and keywords plus) were extracted from all the 1833 publications for Co-occurrence analysis on VOSviewer software. Table 8 shows the top 20 high-frequency keywords. Excluding “ferroptosis” (1,192 times), the most frequently appeared keywords (over 300 times) were “cell death” (434 times), “cancer” (360 times), “apoptosis” (335 times), “iron” (321 times), and “oxidative stress” (321 times). We used Citespace software to cluster the keywords and draw a timeline for keywords after clustering (Figure 5). Total 9 clusters were formed: #0 cell death; #1 prognosis; #2 drug resistance; #3 transporter; #4 photodynamic therapy; #5 mixed-lineage kinase; #6 lipid peroxidation; #7 chemistry; and #8 regulated cell death.

To investigate the research frontiers, we used Citespace software to detect keywords burst (Figure 6). A red line indicates that the corresponding terms frequently appeared during this period, and a blue line represents relatively being used infrequently. Among the top 25 keywords with strong burst strength, “tumor suppression”, with a strength of 9.42, was ranked first, followed by “nonapoptotic cell death” (7.82) and “hepatocellular carcinoma cell” (6.18). Notably, “multidrug resistance” had a duration until 2021 and four keywords (“inhibitor”, “Alzheimer's disease”, “necrosis”, and “mitochondrial permeability transition”) had a duration not less than 5 years.