IMC_pattern

---

This repo is the official code for IMC_pattern

Execution Details

---

Requirements

• Enter the environment and run the following command on the command line:

  pip install -r requirements.txt 

• requirements.txt is a text file containing the dependencies and their version information required by the project.

Execution

---

Datasets

• Melanoma[Link]
• Chang[Link]
• Brain[Link]

Data Preprocessing

• Process data references deal_exp.ipynb deal_mask.ipynb

  # visualization of the cell type
  cd ./data_preprocessing/
  python -u visualize_cell_type.py

Training

python -u train.py --gpu_id=0 --repeat_s=0  --repeat_e=1 --fold_s=0 --fold_e=23 \
--convtype=SAGE --act_op=relu --hd=128 --sag_r=64 --before_layer=1 --after_layer=1 --dropout=0.25 --pool_type=sagpool \
--epoch=45 --early_stop=20 --ckpt_save_epoch=25 --lr=2e-4 --weight_decay=5e-5 --Ks=2 --Ke=10 --K_step=2 --class_num=2 \
--ckpt_path=./checkpoint --res_path=./log_res --gnn_path=./data/melanoma/gnn_data  \ 
--label_path=./data/melanoma/label_and_fold/response_label_dict.pkl --fold_path=./data/melanoma/label_and_fold/leave_one_fold_for_response.pkl

Testing

Obtain the subgraph information of the ROI to be analyzed by testing the best performance model checkpoint saved.

python -u test.py --gpu_id=0 --repeat_s=0  --repeat_e=1 --fold_s=0 --fold_e=23 \
--convtype=SAGE --act_op=relu --hd=128 --sag_r=64 --before_layer=1 --after_layer=1 --dropout=0.25 --pool_type=sagpool \
--lr=2e-4 --weight_decay=5e-5 --Ks=2 --Ke=10 --K_step=2 --class_num=2 \
--ckpt_path=./checkpoint --res_path=./log_res --gnn_path=./data/melanoma/gnn_data  \ 
--label_path=./data/melanoma/label_and_fold/response_label_dict.pkl --fold_path=./data/melanoma/label_and_fold/leave_one_fold_for_response.pkl

Data Postprocessing

• Visualize heatmap:

  cd ./post_processing
  python -u visualize_heatmap.py \
  --graph_path=../data/melanoma/gnn_data \
  --subgraph_path=../log_res/sagpool/Tuning_hd_64_convtype_SAGE_pool_ratio_0.015625_lsim_0.5_act_op_relu_K_2_bl_1_al_1/subgraph \
  --visualize_cell_path=../data/melanoma/vis_cell_type \
  --res_path=../results/sagpool/Tuning_hd_64_convtype_SAGE_pool_ratio_0.015625_lsim_0.5_act_op_relu_K_2_bl_1_al_1 \
  --gpu_id=0 \
  --bg_color=190 \

• UTAG Domain generation

  cd ./post_processing
  python -u gen_utag.py --data_root=../data/melanoma --res_utag_name=utag_results_dist10_leiden.h5ad

• Analysis references Analysis process analysis.ipynb

Reference

To be processed...