Description

Unsupervised machine learning for cell type identification in multiplexed imaging using protein expression and cell neighborhood information without ground truth

Input

Name (Type)
Description
Pattern

meta (map)

Groovy Map containing sample information
e.g. [ id:'test']

img_data (file)

Quantification table with single cells as rows, markers (e.g. CD3 or CD8 but names do not have to match exactly) and X/Y coordinates as columns

*.csv

signature (file)

Signature Matrix containing the definition of cell types according to markers

*.csv

high_thresholds (file)

csv file with user-defined probability high thresholds for anchor cell (row 1) and index cell (row 2) definition

*.csv

low_thresholds (file)

optional csv file with user-defined probability low thresholds for anchor cell (row 1) and index cell (row 2) definition

*.csv

Output

Name (Type)
Description
Pattern

meta (map)

Groovy Map containing sample information
e.g. [ id:'sample1', single_end:false ]

versions (file)

File containing software versions

versions.yml

celltypes (file)

File with final celltype annotations concatenated to the original input quantification, due to the mechanism its non-deterministic

*.csv

quality (file)

File with final calculated marker probabilities for inspection, non-deterministic

*.csv
maintainers
Github user LukasHats
Github user ArozHada
get in touch
Ask a question on Slack Open an issue on GitHub

Tools

celesta
Apache-2.0

Automate unsupervised machine learning cell type identification using both protein expressions and cell spatial neighborhood information

github.com/SchapiroLabor/mcmicro-celesta github.com/SchapiroLabor/mcmicro-celesta/blob/main/README.md https://github.com/plevritis-lab/CELESTA 10.1038/s41592-022-01498-z