DCTD_Team_Aginome-XMU

Team Aginome-XMU submissions to DREAM Challenge Tumor Deconvolution

Installation

1. Clone the GitHub repo.

git clone https://github.com/xmuyulab/DCTD_Team_Aginome-XMU.git

2. Install the following requirements.

python >= 3.6.0
torch 
numpy
pandas
scikit-learn
argh

3. Enter the workspace.

cd DCTD_Team_Aginome-XMU

4. Download pretrained models from here and save to DCTD_Team_Aginome-XMU folder.

   Coarse-grained:

   The Challenge consisted of a “coarse-grained” sub-Challenge, during which participants predicted levels of eight major immune and stromal cell populations (including B cells, CD4+ and CD8+ T cells, NK cells, neutrophils, cells within the monocytic lineage (monocytes/macrophages/dendritic cells), endothelial cells, and fibroblasts).

   Fine-grained:

   The Challenge also consisted of a “fine-grained” sub-Challenge, during which participants further dissected major populations into 14 minor sub-populations according to their functional orientation (e.g., naive B cells, memory B cells, naive CD4 T cells, memory CD4 T cells, naive.CD8 T cells, memory CD8 T cells, regulatory T cells, monocytes, macrophages, myeloid dendritic cells, NK cells, neutrophils, endothelial cells, and fibroblasts).

5. Extract tar.gz file.

tar -zxvf coarse_models.tar.gz
tar -zxvf fine_models.tar.gz

The pretrained models of coarse-grained and fine-grained are saved in sub-folders coarse_models and fine_models respectively.

Note: It takes 20-30 minutes to prepare the environment. Downloading pretrained models take a long time.

Usage

python run_DCTD.py 

    positional arguments:
    {coarse,fine}  Select one of the following sub-commands
        coarse       coarse-grained deconvolution
        fine         fine-grained deconvolution
    
    optional arguments:
        -In IN            Input expression file (with genes specified using HUGO symbols)
        -Out OUT          Output result file
        -scale SCALE      The scale of the expression data (i.e., Log2, Log10, Linear)
        -model MODEL      Trained models directory
        -dataset DATASET  name of test dataset

Input

The input expression data should be a comma separated file with columns associated to sample ID and rows to genes specified using HUGO symbols.

Gene	S1	S2	...
A1BG	0.0038788036146706045	0.0054788910267111225	...
A1BG-AS1	0.0021641861287775527	0.001029015600687667	...
...	...	...	...

Run a demo

Use the following command for coarse-grained deconvolution:

python run_DCTD.py coarse -In demo_data.csv -Out ./prediction.csv -scale Linear -model ./coarse_models/ -dataset demo_data

Or you can use following command for fine-grained deconvolution:

python run_DCTD.py fine -In demo_data.csv -Out ./prediction.csv -scale Linear -model ./fine_models/ -dataset demo_data

Note: It takes only a few seconds to perform cell type proportion prediction.

Expected outcomes

The output file prediction.csv is a comma separated file with 4 columns (cell type, sample ID, predicted proportion and dataset name).

cell.type	sample.id	prediction	dataset.name
CD4.T.cells	S1	0.20734058036020336	demo_data
CD8.T.cells	S1	0.10328292389874755	demo_data
NK.cells	S1	0.0654560242324504	demo_data
...	...	...	demo_data

Note

The coarse-grained and fine-grained models were trained on a gene set that contains 5080 genes. Please check whether these genes exist in your expression profile before performing prediction. If not, we recommend to use DAISM-DNN^XMBD package (https://github.com/xmuyulab/DAISM-XMBD.git) to perform deconvolution by training models from scratch.

Citation

Lin Y, Li H, Xiao X, et al. DAISM-DNN^XMBD: Highly accurate cell type proportion estimation with in silico data augmentation and deep neural networks. Patterns (2022) https://doi.org/10.1016/j.patter.2022.100440

White, B. S., de Reyniès, A., Newman, A. M., Waterfall, J. J., Lamb, A., Petitprez, F., ... & Gentles, A. J. (2022). Community assessment of methods to deconvolve cellular composition from bulk gene expression. bioRxiv. https://www.biorxiv.org/content/10.1101/2022.06.03.494221.abstract