Deep learning-based single-cell analysis pipeline for FLuorescence multiplEX imaging via MELC (Multi-Epitope Ligand Cartography [1]).
Contact: Daria Lazic (daria.lazic@ccri.at)
The pipeline is based on methods for:
-image processing (registration, flat-field correction, retrospective multi-image illumination correction by CIDRE [2])
-cell and nucleus segmentation by Mask R-CNN [3], [4]
-feature extraction
-normalization by negative control secondary antibodies and RESTORE [5]
-single-cell analysis (Cytosplore [6], seaborn)
A compiled release with all necessary dependencies pre-installed is available from dockerhub. Nvidia-docker is required to run the image (for tensorflow-gpu support).
All requirements can be found here.
For interactive and quantitative analysis of single-cell data generated by DeepFLEX, we used:
- Cytosplore: an interactive tool for single-cell analysis (download here)
- Seaborn: a python data visualization library
Navigate to the code folder and run the pipeline.sh script.
Download the MELC imaging data of our 8 samples here.
[1]
Schubert, W. et al. (2006).
Analyzing proteome topology and function by automated multidimensional fluorescence microscopy.
Nature Biotechnology.
[2]
Smith, K. et al. (2015).
CIDRE: An illumination-correction method for optical microscopy.
Nat. Methods, 12, 404-406.
[3]
Kromp, F. et al. (2019).
Evaluation of Deep Learning Architectures for Complex Immunofluorescence Nuclear Image Segmentation.
IEEE.
[4]
Kromp, F. et al. (2020).
An annotated fluorescence image dataset for training nuclear segmentation methods.
Scientific Data, 7, 262.
[5]
Chang, Y.H. et al. (2020).
RESTORE: Robust intEnSiTy nORmalization mEthod for multiplexed imaging.
Commun. Biol., 3, 1-9.
[6]
Höllt, T. et al. (2016).
Cytosplore: Interactive Immune Cell Phenotyping for Large Single-Cell Datasets.
Comput. Graph., 35, 171-180.