一、报告题目：Genome-Wide Multi-Phenotype Association Analysis Method Based on Spline Lasso

二、报告人：孙文渊 博士

三、时 间：2025年10月23日 （星期四）18:30—20:00

四、腾讯会议：678 126 989

报告摘要：While many established variant detection techniques show general predictive strength, they often struggle to precisely pinpoint causal variants within the vast landscape of Single Nucleotide Polymorphisms (SNPs). SNPs intrinsically exhibit structural clustering, organized into predefined groups such as genes or pathways. Furthermore, these SNP groupings frequently correlate strongly with synchronized underlying biological functions.

Inspired by these unique characteristics of SNP data structure, this paper introduces the Multiple Sparse Spline Lasso (MSSL) method, designed for simultaneously selecting both common and low-frequency variants at a group level. Operating under the assumption that complex traits are likely governed by a limited number of single loci within a restricted set of functional genes, the MSSL approach focuses on simultaneously identifying sparse variants and highlighting the relevant functional groups, including the specific loci within them.

Crucially, by leveraging the spline penalty, the method concentrates on the smooth variation of regression coefficients within the same group. We achieve changing-point detection by monitoring the gradual transition of coefficients, rather than relying on abrupt changes. Simulation studies confirm that the MSSL method excels in both feature selection and predictive accuracy, and its advantages are further validated through an application to real-world genetic datasets.

报告人简介：孙文渊，延边大学理学院数学系统计学专业教师，博士毕业于东北师范大学统计学，主要研究方向为高维数据分析，生物统计学。研究论文发表于 BMC bioinformatics、Computers in Biology and Medicine等期刊。

欢迎感兴趣的老师和同学参加！联系人：陶元红