Unsupervised machine learning derived bone phenotypes exhibit differential biomarker responses following acute ballistic loaded exercise.

Authors

Jenna B Goulart, Christopher K Kargl, Adam J Sterczala, Nicole M Sekel, Kelly H Mroz, Mita Lovalekar, Pouneh K Fazeli, Morgan W Bolger, Jane A Cauley, Monica J Hubal, Fabrisia Ambrosio, Brian J Martin, Allison C Bean, Thomas J O'Leary, Julie P Greeves, Kristen J Koltun, Bradley C Nindl

Published in

Bone. Pages 118013. Jul 11, 2026. Epub Jul 11, 2026.

Abstract

Resistance exercise can stimulate new bone formation and result in changes to circulating markers of bone metabolism, but the relationship between the bone metabolic response to resistance exercise and bone morphological phenotypes is unknown. This study compared circulating bone biomarker responses to acute ballistic resistance exercise between groups characterized by bone phenotypes. Fuzzy c-means clustering (n = 287, 47% women) of tibial HR-pQCT parameters and micro finite element analysis (both 4% and 30% sites) determined bone phenotypes. Biomarkers of bone formation (PINP, ALP), resorption (βCTx, TRAP5b), mechanical sensing (sclerostin), and systemic anabolism (IGF-I) were assessed by ELISA before and after an acute ballistic lower body resistance exercise test (AET). DXA assessed body composition. Linear mixed-effects modeling analyzed biomarker responses between clusters, controlling for sex, age, and total lean mass with participants as random intercepts. Clustering revealed two phenotypes (C1 N = 150, 78% women; C2 N = 137, 14% women, p < 0.001), with C2 having wider, denser, and stronger bones with more trabeculae. C2 had higher lean mass (mean difference = 9.1 kg, p < 0.001) than C1. Interaction effects showed IGF-I increased in C2 (p = 0.019) versus no change in C1 (p = 0.999), and TRAP5b decreased to a greater extent in C2 (p < 0.001) compared to C1 (p = 0.029) post-AET. Time effects showed ALP (p = 0.001) and βCTx (p < 0.001) decreased while sclerostin increased (p < 0.001) post-AET overall. Individuals with wider, denser bones exhibit post-exercise biomarker responses potentially conducive to osteogenic adaptation, although the effects on bone structure remain unclear. Unsupervised machine learning derived bone phenotypes provides a novel approach to investigate bone health.

PMID:
42435954
Bibliographic data and abstract were imported from PubMed on 12 Jul 2026.

Read full publication at:
Please sign in to see all details.

Stats

1-terrible, 9-excellent. How would you rate this publication? Sign in in to submit your rating.

Post a comment

You need to be signed in to post comments. You can sign in here.

Comments

There are no comments yet.