Yun Yin

Results To build a predictive model capable of learning cancer-associated signals from routine clinical data, we developed a unified, multimodal framework, Oncoformer, and validated it on a large-scale, multi-cohort dataset (Figure 1). The core of our study is a large-scale discovery cohort, COMPASS-Main, which includes longitudinal data from 2,810,742 individuals.

CORRECTION article Volume 16 - 2025 | https://doi.org/10.3389/fimmu.2025.1773368 This article is part of the Research TopicDistinctive features of tissue-resident and circulating immune memory cellsView all 3 articles The funder “Chengdu Medical Research Project, 2025138” was erroneously omitted. The original version of this article has been updated.

1 Introduction Based on the NCD Risk Factor Collaboration (NCD-RisC) published findings in 2024, it is estimated that more than one billion people in the world were living with obesity, nearly 880 million adults (1). Although weight loss, known to improve metabolic outcomes associated with obesity, is highly recommended for those with obesity, recent studies have reported that most individuals (60%) will regain their lost weight within a few years (2, 3).