Classification and functional characterization of chemokines through machine learning approach

K Vignesh; K Kanagarajadurai

Classification and functional characterization of chemokines through machine learning approach

Author(s):
K Vignesh and K Kanagarajadurai

Abstract:
Chemokines are inflammatory responsible proteins, which mediates varied immune functions like pulling leukocytes towards the inflammatory site, angiogenesis, T-cell differentiation & etc. At present, chemokines are classified in to CC, CXC, CX3C and XC (or just C) based on the patterns of their first two cysteine residues at their N-terminal region. Chemokines bind to G-protein coupled receptors (GPCRs). These GPCRs are classified based on the chemokines that binds to that receptor. More than one ligand binds to a single receptor and similarly, a single ligand binds to more than one receptor. This promiscuous nature of the chemokines and their receptors also extends across different classes (for eg. CCL1 ligand binds to CX3CR1 receptor despite their higher affinity to CCR1). These discrepancies maybe attributed to the classification of Chemokine receptors, which are not based on the receptor properties. So, the current study has been designed to classify the chemokines receptors using Support Vector Machine (SVM) based on the receptor properties exclusively.

There were 19 SVM (Support Vector Machine) models of chemokine receptors were generated to predict any protein sequence to be a chemokines or non-chemokine receptor sequence. Despite that it can also identify its receptor classification. The Relief and mRMR algorithms plays a major role in determining the sensitivity and efficiency of the SVM models. In order to get a better understanding of the SVM output, a phylogenetic tree was constructed using these SVM values. The cluster or a group of receptors based on evolutionary relationship is supported by the work published by other group. The accuracy of the receptor SVM models varies from 83.87% to 100%.

This prediction method of classifying protein sequences by using SVM models, treating each receptor independent of the other and extending it for inferring phylogenetic relationship between them is a novel approach. The achieved accuracy is more, since refinement in accuray was done using Relief and mRMR algorithms. Similar approach may be employed to understand the relationship between protein sequences of interest.

Pages: 3014-3021 | 168 Views 75 Downloads

download (1461KB)

Vol. 10, Issue 12 (2021)

Classification and functional characterization of chemokines through machine learning approach

Important Publications Links

Related Journal Subscription

Important Links