SPiD: An SVM-Based Protein Discriminator for Outer Membrane Proteins
Authors: Babak Alipanahi Ramandi and Ming Li.
Abstract:
Discrimination of Outer Membrane Proteins (OMP) from other types of membrane and globular proteins is an important step in their secondary and tertiary structure prediction. Moreover, a reliable discrimination method can be used for whole genome analysis and hence discovery of new OMPs. In this paper, we propose an SVM-based
protein discriminator for OMPs (SPiD) from other types of proteins, i.e., globular and inner membrane proteins. This approach uses amino acid and amino acid pair composition values, the length of protein sequence, and a newly defined feature called B-barrel score. When applied to a dataset consisting of 1,087 proteins, SPiD achieves an overall accuracy of 96%; to the best of authors' knowledge, this is higher than the accuracy of other previous studies. When SPiD is trained to pick up only outer membrane B-barrels, it reaches an overall accuracy of 99%.