# GeOKG

**Repository Path**: Tomhappy/GeOKG

## Basic Information

- **Project Name**: GeOKG
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: MIT
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2026-03-06
- **Last Updated**: 2026-03-06

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# GeOKG: Geometry-aware representation of Gene Ontology and Genes

### Data Preprocessing
To download and preprocess datasets (GO, GOA, STRING)
```
$ python3 preprocessing/go_string.py
$ python3 preprocessing/goa_string.py
$ filter_gen.py
```

### Embedding GO
To train GeOKG for GO embeddings 
```
$ python3 run.py --dataset GO --model GeOKG_go --rank 200 --batch_size 256 --multi_c
```
Then the evaluation results (GO-level link prediction, link reconstruction, relation-type prediction) are recorded in `evalGO/eval_metrics.tsv`

### Embedding GOA
To train GeOKG for GOA embeddings 
```
$ python3 run.py --dataset GOA --model GeOKG --rank 200 --batch_size 256 --multi_c
```

### PPI Prediction
To evaluate gene embeddings for 3 PPI predictions
* Interaction Binary Prediction
```
$ python evalGene/binary_prediction_NN.py -path evalGO/{date}/GOA/entity_embedding.npy -dset evalGene/GOA_ppi.csv -model result/GeOKG_binary.pth -fout evalGene/binary_output.txt
```
* Interaction Score Prediction
```
$ python evalGene/score_prediction_NN.py -path evalGO/{date}/GOA/entity_embedding.npy -dset evalGene/GOA_score_ppi.csv -model result/GeOKG_score.pth -fout evalGene/score_output.txt
```
* Interaction Type Prediction
```
$ python evalGene/type_prediction_NN.py -path evalGO/{date}/GOA/entity_embedding.npy -dset evalGene/GOA_type_ppi.csv -model result/GeOKG_type.pth -fout evalGene/type_output.txt
```   
   
For the replication of the gene-level experiment results, the pre-trained GOA embeddings are publicly available in the below link:  
[https://zenodo.org/records/15165076]