LCK and HOMER1 gene expression-based classifier distinguish early mycosis fungoides from eczema and psoriasis

This is the repository for the paper: LCK and HOMER1 gene expression-based classifier distinguish early mycosis fungoides from eczema and psoriasis from Meinel et al.

Dependencies

The necessary dependencies can be installed using conda with the given environment.yaml file. conda env create -f environment.yaml

Code

All the scripts are setup with hydra with the corresponding configuration files in the conf directory.

Feature selection with SAFARI

The feature selection method SAFARI can be executed by running Feature_selection_run.py and defining the following arguments in the feature_selection.yaml config:

• counts: summarizedExperiment file from R containing:
  • counts: raw count matrix with genes x samples
  • rowData: containing hugo gene name annotation for ensemble gene names
  • colData: specifying sample by sampleID and disease with diag
• data_split.xlsx with the sheets train and test containing the columns sampleID and diag specifying which samples are used for the feature selection and which ones are not used for unbiased testing later on. All samples can also be used for feature selection.
• d1_nl_path: specifying .rds file with the differential gene expression results comparing condition1 from non-lesional. The indexes are the genes using ensembl_ids as the count matrix and the padj values have to present in one column.
• d2_nl_path: analogous as for d1_nl_path just for the 2nd condition.

The model within SAFARI can be chosen between by setting ffs_model

• logistic
• SVM
• xgboost