Hiring in life sciences? Share your open positions with our professional community. Read more Close

Advertisement

MitoLatentProfiler: ROI-constrained deep learning for mitochondrial phenotyping in human myotubes.

Created on 10 Sep 2026

Authors

Antoine Martin-Tissier, Fahym Bounazou, Aniela Zablocki, Roman Lambert, Guillaume Fargier, Caroline Roelants, Marine Foray, Giulio Morozzi, Marc Liberatore, Erwann Ventre, Joanne Young, Mélanie Flaender, Bianca Freytag

Published in

SLAS discovery : advancing life sciences R & D. Pages 100334. Sep 09, 2026. Epub Sep 09, 2026.

Abstract

Mitochondrial network organization in skeletal muscle reflects metabolic health and is disrupted in primary mitochondrial myopathies, type 2 diabetes/insulin resistance, and age-related functional decline. Studying these disruptions in vitro using mature human myotubes is, however, complicated by the dense, anisotropic architecture of mitochondria, which poses fundamental challenges for automated segmentation and phenotypic classification. Here we present MitoLatentProfiler, a deep learning framework that combines images of micropatterned primary human myotubes with a topology-preserving U-Net and a Classifier U-Net that jointly optimizes segmentation and phenotype classification, structuring the encoder latent space for downstream analysis. Constraining analysis to troponin-positive myotubes excludes confounding signals from neighboring cells. The resulting embeddings organize along two main axes: a morphological axis (fusion/fission balance, shared by TOMM20 and MitoTracker) and a marker-specific axis (associated with bioenergetic insult, MitoTracker only), enabling hierarchical marker-adaptive classification. The segmentation model achieves Dice =0.82 and clDice =0.86. A hierarchical support vector machine classifier reaches macro F1 up to 0.93, outperforming classical morphological descriptors across donors (n=2), markers, and plates not seen during training. Fragmentation, Hypertubulation, and two Damaged phenotype-scores (induced by oligomycin/antimycin and carbonyl cyanide m-chlorophenyl hydrazone) yield Z' factors of 0.54, 0.38, 0.87, and 0.96 respectively, confirming screening applicability for damage and fragmentation readouts. The modular design allows the encoder to serve as a fixed feature extractor: adapting to new phenotypes requires only retraining the lightweight classifier. This pipeline provides a scalable tool for mechanistic studies and compound screening targeting mitochondrial dysfunction in muscle disease.

PMID:
42716305
Bibliographic data and abstract were imported from PubMed on 10 Sep 2026.

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

Advertisement

Stats

  • Community rating n/a 0 votes
  • Reviewers' rating n/a 0 votes
  • Your rating

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

  • Recommendations n/a n/a positive of 0 vote(s)
  • Views 2
  • Comments 0

Recommended by

  • No recommendations yet.

Post a comment

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

Comments

There are no comments yet.

Advertisement