deconvolution_benchmarking - v4.0.0 - Release entire codebase

Version 4.0.0 was released post-publication and publicised entire codebase of the project.

Major changes: - Release all analysis source code in src/, including bulk mixtures simulation, SMOTE, etc - We originally have bisque as logged and scalde. This was later corrected to non-logged, and subsequently non-scaled during the revision process - Batch effect validation experiment: renamed BayesPrism v2 to BayesPrism. Originally we used BayesPrism v1. However, BayesPrism v2 was released during the revision process, hence we swapped BayesPrism v1 for BayesPrism v2 and updated results in the manuscript accordingly

Minor changes: - Tumour purity (with normal epithelial cells) experiment: Update directory in scaden pbs run job - Create params/ directory and added Seurat parameter files (accessed from Wu et al GitHub repository) - Create data/ folder for each experiment and added miniatlas metadata files

Patch changes: None.

Additional notes: - On concatenating AnnData objects: - In later experiments, i.e. 05_external_scrna_validation and 06_batch_effect_validation, concatenation of patient-specific AnnData objects was incorporated into the 91preparetoolspecificdata.ipynb notebook. - In earlier experiments, this step was done in a separate step using Python scripts. On CentOS, these Python scripts could not be run via pbs. We manually set the environment variables and run the Python scripts via command line. This pbs script could behave different with other operating systems. Alternatively, user and extract the code from “08concatenatetestadataobjects.py” and “08concatenatetrainadata_objects.py” directly for this concatenation

- Jupyter Notebook
Published by ktrannt almost 2 years ago

deconvolution_benchmarking - v3.0.1 - Publication

Patch changes: - Update saving of cell-type-specific RMSE values to source data to include matching tumour purity levels with figures, and also in percentage format

- Jupyter Notebook
Published by ktrannt over 2 years ago

deconvolution_benchmarking - v3.0.0 - Publication

Major changes - Added experiment code for batch effect experiment where psedobulks from Bassez et al and Pal et al were deconvolved using single-cell reference profiles from Wu et al - Add visualisation scripts for all main figures and supplementary figures

Minor changes: - Added citation file - Added DOI embedding - Correct paths to CBX singularity container

- Jupyter Notebook
Published by ktrannt over 2 years ago

deconvolution_benchmarking - v2.0.0

• Include external dataset validation with the inclusion of bulk mixtures generated using scRNA-Seq data from Bassez et al and Pal etl
• Replace ABSOLUTE's tumour purity estimates with Consensus Purity Estimates in Aran et al for real bulk validation

- Jupyter Notebook
Published by ktrannt over 2 years ago