meagoodboy / NES

A network-based deep learning methodology for stratification of tumor mutations

paper "A network-based deep learning methodology for stratification of tumor mutations"

Data set

'dataset/Mutation_Individual' directory contains the somatic mutation data for the 15 cancer types.

'dataset/TCGA_Clinical' directory contains the patient clinical data for the 15 cancer types from TCGA.

'dataset/Human Interactome.txt' is the human protein-protein interactome network data.

Code

'struc2vec' directory Contain the protein–protein interactome network embedding method. Each gene generates its own characteristics. python struc2vec/src/main.py --input struc2vec/graph/Human_Interactome.txt --output struc2vec/emb/gene_emb.txt --num-walks 20 --walk-length 80 --window-size 5 --dimensions 2 --OPT1 True --OPT2 True --OPT3 True --until-layer 6

'patient_feature' directory Contain the patient feature construction method. python patient_feature/patient_feature.py

Tutorial

1. To get gene specific expression learned by gene_specific.py.
2. Tumor classification across various cancer types learned by tumour_classification.py.
3. Tumor classification for the specific cancer type learned by specific_patient_classificationA.py.
4. Survival differences between patients learned by patient_cluster.py.

Requirements

This work is tested to work under Python 3.7 The required dependencies for deepDR are gensim, lifelines, pandas, numpy, scipy, and scikit-learn.