Numpy, PyTorch, Sklearn, Scipy, random, seaborn, matplotlib, Math, os
Prediction for each cell type spearately
Processed count data w/ 0~50 cut
/iblm/netapp/home/sherryh/rats/50cut
Or generated from
/iblm/netapp/home/jezhou/make_cnn_training_data/out/pseudobulk
single_cell_binary.ipynb
Note:
Ignore Data Encoding Type 1 and Trainer for Continuous Data and run all other cells.
There are six available models, but using CNN Naive Model would be sufficient.
Prediction for cell types together, including Endothelial, Astrocytes, OPC, InhNeuron, ExNeuron, Microglia, Oligodendrocytes, Sst+.
Merged count data w/ 25~50cut
/iblm/netapp/home/sherryh/binary_data/50cut
/iblm/netapp/home/sherryh/binary_data/25cut
Generated from MACS2
/iblm/netapp/home/sherryh/binary_data/MACS2
Multi_cell.ipynb
Note:
Run all cells and select model of choice, CNN works the best.
Prediction for each cell type spearately
Processed count data
/iblm/netapp/home/sherryh/new_data/zscore/
/iblm/netapp/home/sherryh/new_data/zscore_125/
/iblm/netapp/home/sherryh/new_data/zscore_500/
Or generated from
/iblm/netapp/home/jezhou/make_cnn_training_data/out/pseudobulk
single_cell_continuous.ipynb
Note:
There are seven available models, but only CNN is constructed to handle different datasets. Need to manually change feed forward layer size for other models.
Prediction for cell types together, including Endothelial, Astrocytes, OPC, InhNeuron, ExNeuron, Microglia, Oligodendrocytes, Sst+.
/iblm/netapp/home/sherryh/new_data/zscore/merged_80000_rand.npz
multi_quant.ipynb
Note:
Not recommend for quantitative prediction as it yields much worse results than training inidividually
Prediction for each cell type spearately
Single Cell Data
/iblm/netapp/home/sherryh/expression/full_dataset GTEX data /iblm/netapp/home/sherryh/expression/gtex