ALLIUM PrePro is a library for preprocessing gene expression (GEX) and DNA methylation (DNAm) data to prepare it for prediction using the ALLIUM, a multimodal classifier of molecular subtypes in pediatric acute lymphoblastic leukemia.
Krali, O., Marincevic-Zuniga, Y., Arvidsson, G. et al. Multimodal classification of molecular subtypes in pediatric acute lymphoblastic leukemia. npj Precis. Onc. 7, 131 (2023). https://doi.org/10.1038/s41698-023-00479-5
This repository contains:
- GEX data preprocessing helpers
- metadata generation helpers (use only if ALLIUM has been re-trained with a different gene annotation version)
DNAm preprocessing helpers are still in development.
Conda must be installed on your system.
You will need to activate the allium-prepro conda environment before running any subsequent commands.
Install: conda env create -f environment.yml
Activate: conda activate allium-prepro
Update (after changes to environment.yml): conda env update --file environment.yml --prune
To prepare gene expression for prediction using ALLIUM, you will need a CSV file with raw gene transcript counts. The leftmost column should be HGNC gene symbols or Ensembl identifiers.
Example:
| Sample_1 | Sample_2 | ... | |
|---|---|---|---|
| ETV6 | 10 | 10 | ... |
| SARS1 | 20 | 10 | ... |
| DOC2B | 5 | 10 | ... |
If your data was generated in batches, you can specify a batches file that is formatted as follows:
| id | batch |
|---|---|
| Sample_1 | Batch1 |
| Sample_2 | Batch1 |
| Sample_3 | Batch2 |
| ... | ... |
Alternatively, you may choose to pre-process the batches as separate count files and submit them separately to ALLIUM.
Modify examples/example_gex_prepro.py.
Run it with: python -m examples.example_gex_prepro.
You are now ready to feed your PREFIX.counts.allium.csv file into ALLIUM.
Look at examples/example_ref_prepro.py. Note! The ReferencePreprocessor class only needs to be used in the event that ALLIUM has been re-trained using a different gene annotation version.
Preprocessing for experiments in the MPM Research Group are in the mpm_experiments directory and can be replicated by running python -m mpm_experiments.EXPERIMENT_NAME