Abstract | ||
---|---|---|
Although many machine learning algorithms have been proposed to identify cancer-related genes, their prediction accuracy is still limited due to the complex relationship between cancers and genes. To improve the prediction accuracy, many deep learning based tools have been developed, and they have shown their efficiency to handle complex relationships. To use those tools, a deliberate data representation method is indispensable, since majority tools only take those image-like data as inputs. In this study, we propose a novel network representation method, called Net2Image, to transfer topological networks into image-like datasets. The local topological information of individual vertices from six biomolecular networks and one DNA methylation dataset are encoded as 80 * 6 matrices. They are then employed as inputs to train the model for identifying cancer-related genes using TensorFlow. The numerical experiments show that the proposed method can achieve very high prediction accuracy, which outperforms many existing methods. |
Year | DOI | Venue |
---|---|---|
2017 | 10.1007/978-3-319-59575-7_31 | BIOINFORMATICS RESEARCH AND APPLICATIONS (ISBRA 2017) |
Keywords | Field | DocType |
Deep learning,Biomolecular network,Cancer-related gene,Multiple data integration | Gene,External Data Representation,Vertex (geometry),Topological information,Computer science,Matrix (mathematics),Artificial intelligence,Deep learning,Machine learning | Conference |
Volume | ISSN | Citations |
10330 | 0302-9743 | 0 |
PageRank | References | Authors |
0.34 | 3 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Bolin Chen | 1 | 5 | 2.10 |
Yuqiong Jin | 2 | 0 | 0.34 |
Xuequn Shang | 3 | 99 | 29.07 |