Clinical implications of bone marrow adiposity identified using deep learning, phenome-wide association, and Mendelian randomization in the UK Biobank

• DOI: 10.31219/osf.io/n25tk_v1
• Publication Date: 2025-04-25T17:42:02Z
• Authors: Wei Xu; Ines Mesa-Eguiagaray; David M Morris; Chengjia Wang; Calum Gray; Samuel Sjöström; Giorgos Papanasta...; Sammy Badr; Julien Paccou; Lijuan Wang; Xue Li; Paul R.H.J. Timmers; Maria Timofeeva; Scott I.K. Semple; Tom MacGillivray; Evropi Theodoratou; William Peter Caw...
• Keywords: Phenome; Mendelian Randomization; Mendelian inheritance; Association (psychology)

Bone marrow adipose tissue (BMAT) is a normal feature of mammalian anatomy that increases with ageing and in osteoporosis, type 2 diabetes, other diverse clinical contexts. However, the full scope diseases associated altered bone adiposity remains to be determined, whether BMAT directly contributes human disease unknown. To address these critical gaps knowledge, we previously used deep learning measure fat fraction (BMFF) femoral head, total hip, diaphysis, spine over 44,000 participants UK Biobank, followed by genome-wide association meta-analyses identify genetic architecture BMFF. Here, use data for phenome-wide studies (PheWAS) systematically investigate BMFF at each site. First, conduct PheWAS using measured find it 47 incident across 12 categories. These include not only fracture, but also linked BMAT, such as cardiovascular diseases, several cancers, conditions substantial worldwide burden on health. Sex-stratified PheWASes further demonstrate BMFF-associated differ between males females. We then establish polygenic risk scores (PRSs) PRS-PheWAS Mendelian Randomization explore potential causal outcomes. reveal increased fractures whereas predisposition site positively both diseases. Intriguingly, this negative hip diaphysis. Together, our findings substantially advance understanding impact health promising biomarker and/or therapeutic target improved prevention treatment