Abstract
Protein subcellular localization is an important topic in proteomics since it is related to a protein's overall function, helps in the understanding of metabolic pathways, and in drug design and discovery. In this paper, a basic approximation technique from natural language processing called the linear interpolation smoothing model is applied for predicting protein subcellular localizations. The proposed approach extracts features from syntactical information in protein sequences to build probabilistic profiles using dependency models, which are used in linear interpolation to determine how likely is a sequence to belong to a particular subcellular location. This technique builds a statistical model based on maximum likelihood. It is able to deal effectively with high dimensionality that hinders other traditional classifiers such as Support Vector Machines or k-Nearest Neighbours without sacrificing performance. This approach has been evaluated by predicting subcellular localizations of Gram positive and Gram negative bacterial proteins.
| Original language | English (US) |
|---|---|
| Pages (from-to) | 25-33 |
| Number of pages | 9 |
| Journal | Journal of Theoretical Biology |
| Volume | 386 |
| DOIs | |
| State | Published - Dec 7 2015 |
| Externally published | Yes |
ASJC Scopus subject areas
- Statistics and Probability
- Modeling and Simulation
- Biochemistry, Genetics and Molecular Biology(all)
- Immunology and Microbiology(all)
- Agricultural and Biological Sciences(all)
- Applied Mathematics
Keywords
- Dependency models
- Feature extraction
- Hidden Markov models
- Natural language processing