SeNMo: Self-Normalizing Foundation Model for Enhanced Multi-Omics Data Analysis in Oncology

Overview

SeNMo is a deep learning model designed to enhance the analysis of multi-omics data in oncology. This repository contains the code and instructions for training, fine-tuning, testing, and ensemble testing of the SeNMo model.

Table of Contents

• Requirements
• Running the SeNMo Training
• Running the SeNMo Finetuning
• Running the SeNMo Testing
• Running the SeNMo Ensemble Testing
• Checkpoints
• Embeddings
• Paper
• Future Work

Requirements

Ensure you have all the necessary dependencies installed. You can install them using the requirements.txt file:

pip install -r requirements.txt

Running the SeNMo Training

To train the SeNMo model with the specified parameters, use the following command:

python SeNMo_Training.py \
    --regression True \
    --finetune False  \
    --exp_name surv \
    --act_type None \
    --reg_type all \
    --disease pancancer_combined \
    --task surv \
    --gpu_ids 0 \
    --lr 0.0005811726189177087 \
    --weight_decay 0.005978947728252338 \
    --dropout_rate 0.10583716299176746 \
    --batch_size 256 \
    --dataroot <path_to_data> \
    --checkpoints_dir <path_to_checkpoints> \
    --input_size_omic 80697

Running the SeNMo Finetuning

To finetune the SeNMo model with the specified parameters, use the following command:

python SeNMo_Training.py \
    --regression True \
    --finetune True  \
    --exp_name surv \
    --reg_type all \
    --act_type None \
    --disease pancancer_indl_cancers \
    --task surv \
    --gpu_ids 0 \
    --lr 0.000040189177087 \
    --weight_decay 0.35 \
    --dropout_rate 0.35 \
    --batch_size 16 \
    --niter_decay 8 \
    --dataroot <path_to_data> \
    --checkpoints_dir <path_to_checkpoints> \
    --pretrained_model_dir <path_to_pretrained_model> \
    --cancer 'CPTAC-LUSC' \
    --frozen_layers 0 \
    --input_size_omic 80697

Running the SeNMo Testing

To test the SeNMo model with the specified parameters, use the following command:

python SeNMo_Testing.py \
    --regression True \
    --exp_name surv \
    --reg_type all \
    --act_type None \
    --disease pancancer_combined \
    --task surv \
    --gpu_ids 0 \
    --lr 0.0005811726189177087 \
    --weight_decay 0.005978947728252338 \
    --dropout_rate 0.10583716299176746 \
    --batch_size 1 \
    --niter_decay 8 \
    --dataroot <path_to_data> \
    --checkpoints_dir <path_to_checkpoints> \
    --pretrained_model_dir <path_to_pretrained_model> \
    --frozen_layers 0 \
    --input_size_omic 80697

Running the SeNMo Ensemble Testing

To perform ensemble testing with the SeNMo model, use the following command:

python SeNMo_Ensemble.py \
    --regression True \
    --exp_name surv \
    --reg_type all \
    --act_type None \
    --disease pancancer_combined \
    --task surv \
    --gpu_ids 0 \
    --lr 0.0005811726189177087 \
    --weight_decay 0.005978947728252338 \
    --dropout_rate 0.10583716299176746 \
    --batch_size 1 \
    --dataroot <path_to_data> \
    --checkpoints_dir <path_to_checkpoints> \
    --pretrained_model_dir <path_to_pretrained_model> \
    --input_size_omic 80697

Checkpoints

You can find the pre-trained model checkpoints here, and separately in two parts part-1 and part-2.

Embeddings

Access the TCGA molecular embeddings here, and separately here.

Paper

For detailed information, refer to the arXiv paper.

Future_Work

in progress...creating class packages for CLI pre-processing and generating embeddings at inference time.

• Link to MINDS dataset
• Link to HoneyBee