On January 4, the 18th Academic Symposium on Statistical Methods in Clinical Medical Research was successfully held at Renmin University of China (RUC). The symposium was co-organized by the Center for Applied Statistical Science at RUC, the Department of Biostatistics and Epidemiology at the School of Statistics, RUC, the Institute for Health Big Data Research, RUC, the Beijing Society for Biomedical Statistics and Data Management, and the Health Medical Big Data Branch of the National Industrial Statistics Teaching and Research Association.

Chairperson of the Beijing Society for Biomedical Statistics and Data Management, Prof. Guo Xiuhua from Capital Medical University, delivered the opening remarks. She shared updates on the development of the Beijing Society for Biomedical Statistics and Data Management and emphasized the critical role of statistics in the medical field. She noted that the symposium series on statistical methods in clinical medical research could promote interdisciplinary integration and development, and expressed her hopes for the event's success.

Initiator of the symposium and Prof. Yi Daniao from the School of Statistics, RUC, also gave a speech. She introduced RUC's achievements in public health talent cultivation and future research directions, expressing the hope that the academic exchange platform established by the symposium would actively contribute to the development of China's biomedical statistics.

The morning session of the symposium featured academic presentations chaired by Prof. Li Yang (Dean of the School of Statistics, RUC), Prof. Xu Wangli (Vice Dean of Mingli Academy), and Prof. Zhang Jingxiao (School of Statistics, RUC).

Prof. Wang Qihua from the Academy of Mathematics and Systems Science, Chinese Academy of Sciences (CAS), presented a talk titled “Asymptotic Inference in Decentralized Networks: Penalized Empirical Likelihood with ADMM.” This presentation addressed the needs for large-scale data storage, computation challenges, and privacy protection, leveraging the advantages of empirical likelihood to develop a penalized empirical likelihood inference method for decentralized networks. A theoretical framework was established, and two ADMM-based algorithms for Lagrangian fusion optimization were proposed, opening new pathways for data processing and statistical inference based on large-scale data, aiding in solving practical issues in clinical medicine.

Prof. Xue Fuzhong from Shandong University presented a talk titled “Building a Research-Oriented Clinical Big Data Platform to Promote Hospital IIT Collaborative Innovation.” The presentation highlighted health big data as a foundational strategic national resource critical for advancing medical research, innovation, biopharmaceutical development, and high-end medical device manufacturing. The concept of “open modeling rather than open data” and “model sharing rather than data sharing” was proposed, utilizing federated learning and other big data analysis technologies to address conflicts between privacy protection and data sharing. Finally, Prof. Xue suggested integrating AI-driven disease management methods to form a closed-loop research model and promote medical innovation.

Prof. Hu Yijuan from the Department of Biostatistics at Peking University presented a talk titled “Inferring Microbial Association Networks.” Starting with the basics of microbial omics data and microbial networks, the presentation discussed the challenges posed by the compositional nature and characteristics of sequencing data in microbial network inference. To control false discovery rates, the innovative TestNet testing method was proposed, achieving excellent calibration results.

In the afternoon session, three young scholars and three Ph.D. candidates delivered invited talks, chaired by Assoc. Prof. Wang Chunyan (School of Statistics, RUC) and Ph.D. candidate Xu Shaodong (School of Statistics, RUC).

Assoc. Prof. He Kejun from the Institute for Statistics and Big Data, RUC, presented a talk titled “Modeling Microbial Community Coalescence via Compositional Directed Acyclic Graphical Models.” Addressing the lack of research on microbial community coalescence mechanisms, a novel compositional directed acyclic graphical model was proposed, along with methods for learning graph structures. Two empirical studies provided evidence for microbial community convergence.

Lecturer Zhang Shucong from the School of Statistics, University of International Business and Economics, presented a talk titled “SVEM: Stochastic Variational EM Algorithm for High-Dimensional Multinomial Data.” The presentation addressed challenges in metagenomic sequencing, such as zero-inflation and over-dispersion in bacterial abundance count data. A zero-inflated logistic normal multinomial distribution model was introduced, offering a flexible and effective framework for handling zero-inflation, over-dispersion, and complex dependency structures in microbiome data. To overcome high-dimensional computational challenges, a stochastic variational expectation maximization algorithm was developed, validated through empirical studies.

Lecturer Wang Xinyue from the School of Statistics, RUC, presented a talk titled “Exploring the Use of Artificial Genomes for Genome-Wide Association Studies through the Lens of Utility and Privacy.” The presentation highlighted the potential of collaborative genome-wide association studies (GWAS), hindered by privacy concerns and cumbersome data validation and collaborator selection processes. Advances in generative models offer possibilities for enhancing privacy and accelerating review processes through synthetic genomic datasets resembling real data. The presentation evaluated the ability of deep generative models to generate artificial genomic data for GWAS applications and argued that current privacy measures based on membership inference attacks are insufficient, outlining future directions for effective artificial genome use in GWAS.

Ph.D. candidate Qiao Nan from the School of Statistics, RUC, presented a talk titled “Rank-Based Transfer Learning for High-Dimensional Survival Data with Application to Sepsis Data.” The presentation addressed challenges in estimating key parameters for target populations using multi-source data, proposing a transfer learning method tailored for high-dimensional survival analysis within a transfer model framework. By detecting informative sources similar to target data and using them to improve target models, the method demonstrated practical effectiveness under reasonable conditions through cross-validation and C-index testing. Numerical simulations and MIMIC sepsis data validated the superiority of the proposed method.

Ph.D. candidate Gu Xi from the School of Statistics, RUC, presented a talk titled “Constructing Model Confidence Set via Signal Loss Ratio.” The presentation addressed the uncertainty in model selection in practical applications by introducing a signal loss ratio to measure explanatory power loss between nested models, constructing model confidence sets based on irreducibility and non-expandability hypothesis tests.

Ph.D. candidate Liu Xingwei from the School of Statistics, RUC, presented a talk titled “Ensemble Testing for High-Dimensional Multiple Response Models.” The presentation focused on solving key challenges in high-dimensional multiple response model testing, proposing a multivariate U-statistic-based testing procedure for regression coefficients in multiple response regression models. Under general assumptions, the proposed statistic asymptotically follows a multivariate normal distribution, enabling an ensemble prior-informed test statistic. The integrated testing procedure demonstrated higher power and robustness compared to existing methods.

The symposium provided a vital platform for experts and scholars in biomedical statistics, fostering interdisciplinary collaboration and cross-domain development. Through extensive discussions and collective efforts, the event advanced China's biostatistical landscape, paving the way for future innovations in clinical and health research