Comparing the Performance of BiomimeticPattern Recognition with W8 and SVM onPrediction of Horizontal Gene Transfers inBacteria Genomes
-
Graphical Abstract
-
Abstract
With the accomplishment of sequencing
genomes, lots of homologous genes had been found in dif-
ferent species. These homologous genes are called horizon-
tal transfer genes, in which genetic material is transferred
from the genome of one organism to another. Prediction of
the Horizontal gene transfers (HGT) has important mean-
ing for understanding the genome evolution and estimating
the hereditary material between species. A novel approach
based on Biomimetic pattern recognition (BPR) was pro-
posed to predict the horizontal gene transfers in bacte-
rial genomes. The basis of BPR points to the Principle
of homology-continuity (PHC) that the di®erence between
two samples of the same class must be gradually changed.
The aim of BPR is to ¯nd an optimal covering in the fea-
ture space, which rather than \division" in traditional pat-
tern recognition, emphasizing the \similarity" among ho-
mologous group members. Two neuron models called Hy-
per sausage neuron (HSN) and Ã3-neuron as covering units
in BPR were used to construct the multi-weighted neural
network. The performance of the approach was superior
to that of gene scoring method of 8-nucleotide composition
(W8) and one-class Support vector machine (SVM).
-
-
DownLoad: