Skip to content

Latest commit

 

History

History
68 lines (57 loc) · 3.61 KB

Run Imputation.md

File metadata and controls

68 lines (57 loc) · 3.61 KB

Run Imputation

We use a Linux CentOS 7 machine which has 2 Intel® Xeon® E5-2650 v4 CPUs, 128GB RAM and 1 NVIDIA® Tesla® V100 GPU.

For some datasets which have duplicated gene names, we run a python script in terminal like this

python3 reproducibility/source/loom_rename_duplicated.py \
--input=matrix.loom

The renamed loom-formatted file will be generated in the same folder as matrix.loom.

SAVER

R --slave < "./reproducibility/source/Running Scripts for Other Methods/SAVER.r" \
--args /home/yuanhao/github_repositories/DISC/reproducibility/data/MELANOMA/raw.loom 16 10

scVI

python3 "./reproducibility/source/Running Scripts for Other Methods/scVI.py" \
--loom=/home/yuanhao/github_repositories/DISC/reproducibility/data/MELANOMA/raw.loom \
--min-expressed-cell=10

MAGIC

python3 "./reproducibility/source/Running Scripts for Other Methods/MAGIC.py" \
--loom=/home/yuanhao/github_repositories/DISC/reproducibility/data/MELANOMA/raw.loom \
--min-expressed-cell=10

DCA

python3 "./reproducibility/source/Running Scripts for Other Methods/DCA.py" \
--loom=/home/yuanhao/github_repositories/DISC/reproducibility/data/MELANOMA/raw.loom \
--min-expressed-cell=10

scScope

python3 "./reproducibility/source/Running Scripts for Other Methods/scScope.py" \
--loom=/home/yuanhao/github_repositories/DISC/reproducibility/data/MELANOMA/raw.loom \
--min-expressed-cell=10

DeepImpute

python3 "./reproducibility/source/Running Scripts for Other Methods/DeepImpute.py" \
--loom=/home/yuanhao/github_repositories/DISC/reproducibility/data/MELANOMA/raw.loom \
--min-expressed-cell=10

VIPER

R --slave < "./reproducibility/source/Running Scripts for Other Methods/VIPER.r" \
--args /home/yuanhao/github_repositories/DISC/reproducibility/data/MELANOMA/raw.loom gene 10

scImpute

R --slave < "./reproducibility/source/Running Scripts for Other Methods/scImpute.r" \
--args /home/yuanhao/github_repositories/DISC/reproducibility/data/MELANOMA/raw.loom 16 10

Here, all results will be saved in /home/yuanhao/data/fn/melanoma/imputation with genes filtered. We can easily run

python3 reproducibility/source/resume_dim.py \
--raw-loom=/home/yuanhao/github_repositories/DISC/reproducibility/data/MELANOMA/raw.loom \
--impute-h5=output_file

and get output_file_resume_dim.loom with complete dimension (only genes as we don't resume cells here) as /home/yuanhao/github_repositories/DISC/reproducibility/data/MELANOMA/raw.loom.

DISC

For DISC, the imputation result (DISC outputs a matrix of the same dimension with input.loom, in which genes selected by "gene selection" are updated using DISC imputaion result) is saved in out_dir/result/imputation.loom, the low dimensional (default 50) of DISC imputaion result is saved in out_dir/result/feature.loom.

disc \
--dataset=/home/yuanhao/github_repositories/DISC/reproducibility/data/MELANOMA/raw.loom \
--out-dir=/home/yuanhao/DISC_imputation_result/melanoma \
--min-expressed-cell=10 \
--library-size-factor=median

If you want to use a specific model trained by DISC for inference, please use this following code:

disc \
--dataset=/home/yuanhao/github_repositories/DISC/reproducibility/data/MELANOMA/raw.loom \
--out-dir=/home/yuanhao/DISC_imputation_result/melanoma_inference \
--pretrained-model=/home/yuanhao/DISC_imputation_result/melanoma/models/XXX.pb \
--training=0 \
--min-expressed-cell=10 \
--library-size-factor=median

Please note that using --training=0 here, otherwise DISC will use this model for fine-tune training.