Liang Jiaqi, Xue Zhao, Zhou Wenchao, Guo Xiangjie, Wen Yalu
Academy of Medical Sciences, Shanxi Medical University, Taiyuan, Shanxi, China.
Department of Health Statistics, School of Public Health, Shanxi Medical University, Taiyuan, Shanxi, China.
Front Genet. 2025 Jun 10;16:1538544. doi: 10.3389/fgene.2025.1538544. eCollection 2025.
Correlated phenotypes may have both shared and unique causal factors, and jointly modeling these phenotypes can enhance prediction performance by enabling efficient information transfer.
We propose an auto-branch multi-task learning model within a deep learning framework for the simultaneous prediction of multiple correlated phenotypes. This model dynamically branches from a hard parameter sharing structure to prevent negative information transfer, ensuring that parameter sharing among phenotypes is beneficial.
Through simulation studies and analysis of seven Alzheimer's disease-related phenotypes, our method consistently outperformed Multi-Lasso model, single-task learning approaches, and commonly used hard parameter sharing models with predefine shared layers. These analyses also reveal that while genetic contributions across phenotypes are similar, the relative influence of each genetic factor varies substantially among phenotypes.
相关的表型可能具有共同的和独特的因果因素，对这些表型进行联合建模可以通过实现有效的信息传递来提高预测性能。
我们在深度学习框架内提出了一种自动分支多任务学习模型，用于同时预测多个相关表型。该模型从硬参数共享结构动态分支，以防止负面信息传递，确保表型之间的参数共享是有益的。
通过模拟研究和对七种与阿尔茨海默病相关的表型的分析，我们的方法始终优于多套索模型、单任务学习方法以及具有预定义共享层的常用硬参数共享模型。这些分析还表明，虽然各表型间的遗传贡献相似，但每个遗传因素在各表型间的相对影响差异很大。
