Catchxu/Docs

Docs: Detecting outlier cells and subtypes for single-cell transcriptomics via adversarial training.

Docs: Detecting outlier cells and subtypes for single-cell transcriptomics via adversarial training

We proposed a GAN-based model named Docs (Detecting outlier cells and subtypes). This approach employs a pipeline to integrate multi-task generative adversarial networks for detecting anomaly cells and the subtypes of these cells in single-cell transcriptomics (scRNA-seq and scATAC-seq data).

Download and unzip Datasets

• Download needed datasets from this link: ODDatasets.
• Unzip the ODDatasets.zip file.
• All the datasets are stored as 'h5ad', and can be read by Scanpy or docs.read.

scRNA-seq PBMC Dataset Information

Dataset	Cells	Genes	Sparsity ratio(%)	Anomaly ratio(%)	Anomaly Type
PBMC_3000_ref	4698	3000	96.82
PBMC_3000_B	3684	3000	96.89	13.14	B cells
PBMC_3000_NK	3253	3000	96.48	12.73	NK cells
PBMC_6000_ref	4698	6000	97.52
PBMC_6000_B	3684	6000	97.55	13.14	B cells
PBMC_6000_NK	3253	6000	97.27	12.73	NK cells
PBMC_full_ref	4698	32738	98.82
PBMC_full_B	3684	32738	98.83	13.14	B cells
PBMC_full_NK	3253	32738	98.71	12.73	NK cells

scRNA-seq Lung Cancer Dataset Information

Dataset	Cells	Genes	Sparsity ratio(%)	Anomaly ratio(%)	Anomaly Type
Cancer_3000_ref	8104	3000	94.94
Cancer_3000_EI	7721	3000	96.08	50.03	Epithelial & Immune Tumor
Cancer_3000_ES	4950	3000	94.86	58.12	Epithelial & Stromal Tumor
Cancer_6000_ref	8104	6000	94.85
Cancer_6000_EI	7721	6000	95.99	50.03	Epithelial & Immune Tumor
Cancer_6000_ES	4950	6000	94.32	58.12	Epithelial & Stromal Tumor
Cancer_full_ref	8104	33538	92.90
Cancer_full_EI	7721	33538	94.68	50.03	Epithelial & Immune Tumor
Cancer_full_ES	4950	33538	93.15	58.12	Epithelial & Stromal Tumor

scATAC-seq TME Dataset Information

Dataset	Cells	Genes	Sparsity ratio(%)	Anomaly ratio(%)	Anomaly Type
TME_3000_ref	3559	3000	78.14
TME_3000_Tumor	2968	3000	65.46	60.24	Tumor(pre)
TME_6000_ref	3559	6000	76.61
TME_6000_Tumor	2968	6000	64.70	60.24	Tumor(pre)
TME_full_ref	3559	23127	71.53
TME_full_Tumor	2968	23127	61.02	60.24	Tumor(pre)