Virtual biomarkers: predicting immune status using label-free holotomography of individual human monocytes and machine learning analysis
Lee, M.; Kim, G.; Lee, M. S.; Shin, J. W.; Lee, J. H.; Ryu, D. H.; Kim, Y. S.; Chung, Y.; Kim, K. S.; Park, Y.
Show abstract
Precise evaluation of immune status is critical for managing diseases such as sepsis, in which the immune system transitions between hyper-inflammatory and immune-suppressed states. However, current biomarkers are limited by low specificity and time-consuming protocols. Here, we present a label-free, imaging-based framework for single-cell immune profiling of human monocytes using three-dimensional holotomography (HT) and deep learning. HT captures subcellular refractive index (RI) distributions of live, unlabeled cells, enabling quantitative extraction of morphological and biophysical features. Using an in vitro LPS-based model, we classified three functional immune states--control, hyper-inflammation, and immune suppression--based on 4,059 holotomograms from 11 donors. Immune state transitions were associated with significant changes in cell volume, surface area, RI variability, and the abundance of lipid droplets. A 3D convolutional neural network trained on HT images achieved 83.7% accuracy for single-cell predictions, increasing to 99.9% with ensemble averaging. This study establishes HT as a scalable, label-free platform for real-time immune monitoring and introduces subcellular RI features as robust correlates of immune dysregulation.
Matching journals
The top 9 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- TrimNN: Characterizing cellular community motifs for studying multicellular topological organization in complex tissues 95%
- De novo analysis of bulk RNA-seq data at spatially resolved single-cell resolution 95%
- ROGUE: an entropy-based universal metric for assessing the purity of single cell population 94%
Similar papers in this journal
- Predicting MammaPrint Recurrence Risk from Breast Cancer Pathological Images Using a Weakly Supervised Transformer 94%
- ProT-Diff: A Modularized and Efficient Approach to De Novo Generation of Antimicrobial Peptide Sequences through Integration of Protein Language Model and Diffusion Model 94%
- High-throughput, label-free and slide-free histological imaging by computational microscopy and unsupervised learning 93%
Similar papers in this journal
Similar papers in this journal
- Transformer-based spatial-temporal detection of apoptotic cell death in live-cell imaging. 94%
- Modular DNA Barcoding of Nanobodies Enables Multiplexed in situ Protein Imaging and High-throughput Biomolecule Detection 93%
- Retinal microvascular and neuronal pathologies probed in vivo by adaptive optical two-photon fluorescence microscopy 93%
Similar papers in this journal
- S3-CIMA: Supervised spatial single-cell image analysis for the identification of disease-associated cell type compositions in tissue 94%
- Essential Regression - a generalizable framework for inferring causal latent factors from multi-omic human datasets 94%
- DAISM-DNNXMBD: Highly accurate cell type proportion estimation with in silico data augmentation and deep neural networks 92%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.