Abstract

Phenotypic plasticity is essential to the immune system, yet the factors that shape its cell morphologies are not fully understood. Here, we comprehensively analyze immune cell morphology across a human cohort by single-round multiplexed immunofluorescence, automated microscopy, and deep learning. Using the uncertainty of convolutional neural networks to cluster the morphologies of eight distinct immune cell classes, we find that the resulting cell morphology maps are influenced by the donor’s age, gender, and blood pressure, revealing distinct polarization and activation-associated phenotypes across immune cell classes. We further associate gene expression to T-cell morphology based on their joint donor variability, and validate an inflammation-associated uropod-positive T-cell morphology, and an age-associated loss of mitochondria in CD4+ T-cells. Taken together, we show that immune cell morphologies encode both molecular and personal health information, opening new perspectives into the deep immune phenotyping of individual people in health and disease.

Downloads

Preamble

The code and data presented here are currently for reviewer purpose only.

Code

The Matlab code to perform Local Enrichment Analysis (LEA) using K Nearest Neighbors can be downloaded here as ZIP file: knnlea.zip (2KB). Furthermore the annotated source code can be viewed here.

Datasets

The single-cell CNN training and test datasets generated in this study are available as either a ZIP or TAR file, both here as well as at the ETHZ Research Collection. It is recommended to use download resume-capable software (such as most modern browsers), in case the large downloads get interrupted:
DeepPhenotype_PBMC_ImageSet_YSeverin.zip (1.76GB)
DeepPhenotype_PBMC_ImageSet_YSeverin.tar.gz (1.81GB)

We occasionally see download failures on these large files due to network errors. If this happens, the download can be finished by 'resuming' it.

---

Resume download on Google Chrome:
Resume download on Mozilla Firefox: