Metabolic Networks Features
By Ari Kusumastuti
Metabolic Networks Features
Including all features of T2DM extracted from GTEx patient data and specific tissues (T2DM & Liver - Pancreas).
Met2MetGraph
This repository contains metabolic network features extracted from sample patient data, including Type 2 Diabetes Mellitus (T2DM)-related tissue data and liver-pancreas tissue data. The repository is divided into two main sections:
- T2DM Features
- Liver Features
Each section includes features organized into the following schemes:
- MetGraph: Original features derived from metabolic networks.
- NDD (Node Distance Dependency): Features based on the distance relationships among metabolites.
- TM1: First-order random walk features originating from specific metabolites and their immediate neighbors.
- TM2: Second-order random walk features capturing broader metabolite neighborhood structures.
- NDD+TM1: Combined features incorporating both NDD and first-order random walks.
- NDD+TM1+TM2: Integrated features combining NDD, TM1, and TM2.
Graph Embedding
We employ several graph embedding algorithms, including:
- Graph2Vec
- GL2Vec
- FeatherGraph
- Netpro2Vec
These embeddings are generated using MetGraph features as the primary input source. The features within are generated by combining four patient groups (100, 200, 300, 400), two threshold settings (1_80 and 2_80), three aggregation schemes (MeanSum, MinMax, MinSum), and three embedding dimensions (64, 128, 196). So, we have 72 features within FeatherGraph, GL2Vec, Graph2Vec, and several Netpro2vec in total.