Related Experiment Videos
Mapping local cluster separation, global community structure, and leadership concentration among the 2025 top 2%
Yu-An Wang1, Po-Chih Lai2, Feng-Jie Lai1
1Department of Dermatology, Chi Mei Medical Center, Tainan, Taiwan.
Background:
The Stanford/Ioannidis top 2% scientists database provides standardized citation-based indicators for identifying influential researchers, but conventional rankings offer limited information about the structural organization of research communities and collaboration patterns. This study aimed to develop and demonstrate an integrated bibliometric reporting framework for evaluating local cluster separation, global community structure, and leadership concentration among the 2025 top 2% dermatology scientists.
Methods:
Data for 463 dermatology scientists from 35 countries and 360 institutions were analyzed. Country-, institution-, and coauthor-based networks were constructed and evaluated using 3 complementary metrics: silhouette scores (SS) for local cluster separation, normalized modularity (Q*) for global community structure, and the absolute advantage coefficient (AAC) for leadership concentration. Network maps, silhouette plots, profile summaries, and Kano-style visualizations were used to facilitate interpretation. Dirk Schadendorf and Michael R. Hamblin were selected as illustrative scholar-level cases based on their citation and field-ranking prominence.
Results:
The top 100 high-frequency country and institutional nodes formed 15 clusters. The United States, Germany, the United Kingdom, Australia, and Sweden were the most prominent countries, with Harvard Medical School, Charité-Universitätsmedizin Berlin, and the University of California, San Francisco emerging as major institutional hubs. Schadendorf was the most highly cited scholar, whereas Hamblin ranked highest within dermatology. Their coauthor networks showed similar local and global cluster characteristics but different leadership-concentration patterns: Schadendorf had SS = 0.70, Q* = 0.75, and AAC = 0.66, whereas Hamblin had SS = 0.71, Q* = 0.74, and AAC = 0.76. Under the proposed AAC classification, these values indicate more distributed multicore collaboration for Schadendorf and greater single-core leadership concentration for Hamblin.
Conclusion:
Integrating SS, Q*, and AAC provides complementary information on local cluster separation, global community organization, and leadership concentration that cannot be obtained from citation rankings alone. The framework offers a reproducible and interpretable approach for research mapping, comparative assessment, and longitudinal monitoring across scientific disciplines. AAC should be interpreted as an indicator of collaboration leadership concentration rather than direct evidence of mentorship.
