On August 2, statisticians from Vanderbilt University and three major Taiwanese research institutions—Academia Sinica, National Cheng Kung University, and National Tsing Hua University—convened at Vanderbilt’s Carmichael College for “Next-Generation Statistics: From Theory to Practice in a Data-Rich Era,” a half-day symposium. The event, which was free and open to the public, drew faculty, staff, and student scientists from multiple departments, as well as visitors arriving in Nashville for the 2025 Joint Statistical Meetings (the largest annual gathering of statisticians in North America). It is the first of an international multi-year series co-organized by Vanderbilt’s Department of Biostatistics and Academia Sinica, with plans to conduct the next symposium in Taipei, Taiwan, in 2027.
In a ceremony before the start of the symposium, Department of Biostatistics chair Yu Shyr, PhD, FASA, FAAAS, FAACR, and Academia Sinica’s Institute of Statistical Science director Hsin-Chou Wang, PhD, signed a Memorandum of Understanding to formalize plans for joint initiatives between their institutions. In addition to the symposium series, they envision facilitating visiting scholar exchanges and research collaborations that may utilize resources such as the Alliance of Genomic Discovery (AGD) and Taiwan Biobank databases.
The first session of the symposium was dedicated to work involving statistical genetics, the microbiome, and bioinformatics. Academia Sinica’s Hsuan-Yu Chen, PhD, began the session by detailing how multi-omics profiling and modeling can improve our ability to personalize the treatment of individuals with cancer, especially when environmental carcinogens (e.g., air pollution) may be a factor. His presentation included work published just two days earlier in Cancer Cell. The second talk, by Siyuan Ma, PhD, assistant professor of biostatistics, provided an overview of genomic large language models (LLMs), followed by a closer look at how high-quality gene clustering can boost statistical power and assist in addressing major issues in biomedical research, such as unannotated genes. Vanderbilt doctoral student Yunbi Nam and master’s student Xuexin Li are deeply engaged with this topic, with Nam co-first-authoring a 2024 Frontiers in Genetics paper on advances in LLMs for studying the microbiome.
The third presenter, An-Shu Tai, PhD, serves on the faculty of National Tsing Hua University, with single-cell annotation and causal inference among his research interests. In his talk, he introduced a new statistical method, Med_MR_Egger, which utilizes the landmark Mendelian randomization tool to extend the framework of successful mediation analysis. The new method can help researchers more accurately analyze the impact of lifestyle and environmental factors on health, such as connection between exposure to heavy metals and lung impairment. A panel review of the first three presentations rounded out this first session, moderated by Qingxia Chen, PhD, professor of biostatistics, biomedical informatics, and ophthalmology and visual sciences; the discussants included Ran Tao, PhD, associate professor of biostatistics; Quanhu Sheng, PhD, associate professor of biostatistics; Shilin Zhao, assistant professor of biostatistics; and Chih-Yuan Hsu, research assistant professor of biostatistics.
The second session of the symposium featured innovative methods in bioinformatics and longitudinal data. It began with Qi Liu, PhD, professor of biostatistics and biomedical informatics, reviewing several existing methods for spatial clustering and then presenting a robust new not-yet-published method, MEcell, that offers microenvironment-aware spatial modeling of cell identities. The second speaker, Academia Sinica’s Frederick Kin Hing Phoa, PhD, likewise presented a patent-pending proprietary method for identifying new biomarkers via proteomic analyses that may simultaneously increase the speed and decrease the cost of diagnosing conditions such as Parkinson’s disease.
Bryan Shepherd, PhD, FASA, professor of biostatistics and biomedical informatics, has devoted a substantial part of his career to addressing researchers’ “naïve” handling of biomarker values that fall below detection limits. His talk highlighted the work of alumna Yuqi Tian (PhD 2022), who first-authored several papers on cumulative probability methods that were accepted by top peer-reviewed journals. Andrew Spieker, PhD, associate professor of biostatistics, delivered a dynamic presentation on using a semiparametric approach to address variations in AUC (Area Under the Curve) estimates caused by idiosyncratic practices across different labs, building on work first-authored by Nathaniel Dowd (MS 2024), now a research statistician at Vanderbilt’s Frist Center for Autism and Innovation. The discussants for this second session included Leena Choi, PhD, professor of biostatistics and biomedical informatics, and Kuo-Jung Lee, PhD, a former visiting scholar at Vanderbilt who was named chair of National Cheng Kung University’s Department of Statistics in July.
The focus of the symposium’s final session was statistical methods in imaging. It began with Simon Vandekar, PhD, associate professor of biostatistics, who presented the recommendations for improving neuroimaging study design and replicability published last fall in Nature, led by recent graduate Kaidi Kang (PhD 2025), with current PhD candidates Megan Jones and Jiangmei Xiong among its co-authors. Academia Sinica research fellow I-Ping Tu, PhD, spoke on traditional and ground-breaking methods in electron microscopy, comparing different approaches to mapping proteins such as apoferritin and proposing a hierarchical Bayesian model for better estimates. Hakmook Kang, PhD, associate professor of biostatistics, completed this third set of talks by discussing structural and functional connectivity measures and demonstrating how the GPU (Graphics Processing Unit)-enhanced Gaussian Process developed by his team in the course of conducting autism research can dramatically speed up computing time (completing processes up to 240 times faster than older methods). Kang’s collaborators include Vanderbilt data science students Margo Kim and Yuting Mei, staff biostatistician Ke Xu, MS, and adjoint associate professor Chris Fonnesbeck, PhD.
The discussants for this final session included Jinyuan Liu, PhD, assistant professor of biostatistics and psychiatry, and Chun-houh Chen, adjoint professor of biostatistics at Vanderbilt University Medical Center and secretary-general of Academia Sinica. In his response, Chen observed that he is scheduled to retire in 2026 but found the symposium so inspiring that he is tempted to apply for an extension, to remain involved in staging the 2027 edition. Vanderbilt chair Shyr likewise congratulated all the participants on the high quality of the talks and panels. After the gathering was adjourned, the Department hosted a lunch at its 2525 West End Avenue headquarters for the presenters, followed by a tour.