This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License
|
||||||||
|
Paper Details
Paper Title
Recognition of specific Contact Frequencies Signatures in Protein Structural Classes by Support Vector Machine Learning
Authors
  Fatin Jannus ,  Hilario Ramírez-Rodrigo
Abstract
While regular assigning of major structural classes (all-α, all-β, α+ β and α/ β) are actually used by the most popular classification systems, we still lack of an in-deep understanding about the underlying structural features, particularly, at the level of their residue contacts profiles. Here we describe a study that makes use of Support Vector based Machine Learning algorithms (SVM) to see if these categories can be distinguished in this context or not. To achieve this goal we have developed different learning models that were trained with 400-dimensional contact frequencies vectors sets, previously calculated from a non-redundant sample of 2484 proteins structures. We have built binary and multi-class classification models with mean accuracies of 82% and 60%, respectively. Using these models, it has been possible to binary classify any two structural classes sharing few mixed secondary structures (such as all-α and all-β proteins) with as high as 87% accuracy. This value decreased to 82% if the structural classes share mixed secondary structures to a large extension (like α+ β and α/β). These results are consistent with the existence of differentiated contact frequency profiles for mainly-alpha and mainly-beta protein classes and suggest that α/β protein class could also have a mild specific signature in terms of residue-residue contacts, whereas α+ β class could possibly be discarded with this regard, lacking of specific pattern of contact frequencies. This last finding opens the question of whether α+ β class needs to be redefined to improve coherence of this protein taxonomy.
Keywords- Secondary structure, Residue contact, Machine Learning, Alpha, Beta.
Publication Details
Unique Identification Number - IJEDR1701097Page Number(s) - 626-631Pubished in - Volume 5 | Issue 1 | March 2017DOI (Digital Object Identifier) -    Publisher - IJEDR (ISSN - 2321-9939)
Cite this Article
  Fatin Jannus ,  Hilario Ramírez-Rodrigo ,   "Recognition of specific Contact Frequencies Signatures in Protein Structural Classes by Support Vector Machine Learning", International Journal of Engineering Development and Research (IJEDR), ISSN:2321-9939, Volume.5, Issue 1, pp.626-631, March 2017, Available at :http://www.ijedr.org/papers/IJEDR1701097.pdf
Article Preview
|
|
||||||
|