a phenotype-driven model for gene prioritization.The heterogeneous graph embedding algorithm is used to learn the embedding representation of heterogeneous graph nodes, and the score of candidate causal genes is calculated according to the embedding, so as to prioritize.
The data used in the training model are placed in ./data/ .The edge relation, weight and trained embedding information have been encapsulated by pickle module. In addition, the test data used are also included. In ./models/, it includes the method of constructing graphs, node coding method, training method, test code, etc
Randomly select 999 genes from your own genetic range, and add pathogenic genes to form a gene set of size 1000. In this way, create ten gene sets and calculate the ranking of pathogenic genes in the corresponding gene set each time. Take the median of the ten gene sets as the final ranking
In folder ./models/prioritize/Het2Gene/ ,run the following command can use Het2Gene:
python het2gene.py --hps [parameter] --out_dir [-options][parameter] --topn [-options][parameter]
hps: Abnormal phenotype set from patients，required
out_dir: Directory for outputting results，default RankResult
topn: Output candidate genes in the top n,default 1000
An example command:
python het2gene.py --hps HP:0000573,HP:0001102,HP:0003115,HP:0001681,HP:0008067,HP:0004417 --out_dir myResult --topn 5
Output Example：
Rank EntrezID Symbol Score
1 368 ABCC6 11.56
2 64132 XYLT2 1.983
3 5167 ENPP1 -0.815
4 64131 XYLT1 -0.936
5 4000 LMNA -3.271
CADA:https://cada.gene-talk.de/webservice
AMELIE:https://amelie.stanford.edu
Phen2Gene:https://phen2gene.wglab.org
GADO:https://genenetwork.nl/gado
PhenoApt:[https://www.phenoapt.org/]
See the LICENSE file for license rights and limitations