Paper
A Method for Predicting Protein Functions Based on the Protein Ray-Based Descriptor
- Authors:
- Georgina Mirceva; Slobodan Kalajdziski; Andrea Kulakov
- Abstract
- Protein molecules are involved in many processes in the cells of living organisms, thus the knowledge about their functions in these processes is very important. There are various methods for annotating protein structures. The databases that hold knowledge about the proteins’ annotations cover a portion of the known protein structures. Therefore, one of the most attractive issues in bioinformatics community is automatic discovery of protein functions. In our previous research we proposed several methods for protein tertiary structure retrieval. In this paper we introduce a method for determining protein functions based on our protein ray-based descriptor. First, the query structure is compared with the training structures. Then, the annotations of the query are determined based on the annotations of its nearest neighbours. The results show that our method achieves 0.82 micro precision and 0.74 micro recall. We also compare the annotations inferred by our methods with the annotations with evidence code IEA, and against the annotations with evidence codes different from IEA. Similarly, we compare the annotations inferred by our method with the annotations stored in InterPro and UniProtKB databases.
- Keywords
- Protein Structure; Protein Function; Protein Ray-Based Descriptor; Nearest Neighbours; Multi-Label Learning
- StartPage
- 26
- EndPage
- 33
- Doi