Preprocessed dataset for scClone2DR analysis in acute myeloid leukemia, derived from the Tumor Profiler study.

This dataset corresponds to the version of the data prepared for applying the scClone2DR method to the acute myeloid leukemia (AML) cohort of the Tumor Profiler (TuPro) study. The original AML data was published in

Wegmann, R., Bonilla, X., Casanova, R. et al. Single-cell landscape of innate and acquired drug resistance in acute myeloid leukemia. Nat Commun 15, 9402 (2024). https://doi.org/10.1038/s41467-024-53535-4

Data preprocessed for scClone2DR: 

This folder is structured as follows:

info_cohort.csv: Provides the mapping from sample IDs to patient IDs

pharmacoscopy.csv: Pharmacoscopy data for the different samples obtained from the FastDrug measurements.

Condition: drug and concentration used

Drug: drug tested in the well

Concentration: drug concentration

Well_position_1: x position of the well on the plate

Well_position_2: y position of the well on the plate

Number_tumor_cells: total number of tumor cells

Number_all_cells: total number of surviving cells

SampleID: sample identifier

metacells_geneOncoKB_phenograph/: This folder contains the preprocessed metadata and RNA data for scClone2DR. For the AML data, metacells were clustered using the Phenograph algorithm, leading to different clones (i.e. groups) of metacells. RNA features at the metacell level represent raw counts for the selected gene set from the OncoKB gene list.

clone_infos.csv: Associates each clone ID with its corresponding biological information, namely:

clonecategory: Classifies each clone as either healthy or tumor based on whether it corresponds to non-malignant or tumor metacells 

clonelabel: Classifies each clone as healthy, putative, or tumor based on whether it corresponds to non-malignant, putative tumor, or tumor metacells.

clonetype_{sampleID}: Provides, for each sample, the dominant cell type among the metacells belonging to that clone.

sample2data/: Folder containing for each sample a file with the preprocessed RNA data for all metacells and their clone assignment. Namely for each sample we have:

cell_id: metacell identifier

dim_{i}_{gene}: For i ranging from 1 to 867, provides the raw metacell count for the specified gene

cloneID: Clone ID of the metacell (obtained from the phenograph algorithm)

clonecategory: Clone category of the clone corresponding to the metacell (healthy or tumor)

clonelabel: Clone label of the clone corresponding to the metacell (healthy, putative or tumor)

clonetype: Dominant cell types in the clone to which the metacell belongs

celltype: Cell type of the metacell

metacells_hallmarks_phenograph/: This folder is analogous to metacells_geneOncoKB_phenograph/, but the metacell-level features are derived from pathway-level GSVA scores (using the Hallmark gene sets) rather than from raw counts.

Model weights: 

We provide the learned parameters of the scClone2DR model trained on the AML data.