Tweet
Comput Biol Chem. 2018 Dec 1. pii: S1476-9271(18)30793-X. doi: 10.1016/j.compbiolchem.2018.11.029. [Epub ahead of print]
Predicting interspecies transmission of avian influenza virus based on wavelet packet decomposition.
Qiang X1, Kou Z2.
Author information
Abstract
Using wavelet packet decomposition, the energy coefficients in the fifth level of viral protein sequences were achieved to predict interspecies transmission. Since avian-origin influenza viruses could have high sequence similarities with human-origin avian influenza virus and could have the phenotype of interspecies transmission, viral data should be filtered to prevent the misconduct of feature selection and false performance of predicting models. Considering the balance of data size, the empirical cut-off value 97% was used to screen avian-origin influenza virus with high sequence similarity. The excellent performances of cross validation show that the SVM model has the best capability of predicting transmission and evaluating the contribution of five amino acid factors. The robust model was finally used to evaluate the filtered data of avian-origin virus and the results confirmed that double check for ambiguous phenotype of avian-origin virus with high sequence similarity was necessary and part of them have the ability to across species barriers.
KEYWORDS:
Avian influenza virus; Interspecies transmission; Machine learning; Wavelet package
PMID: 30528510 DOI: 10.1016/j.compbiolchem.2018.11.029
Predicting interspecies transmission of avian influenza virus based on wavelet packet decomposition.
Qiang X1, Kou Z2.
Author information
Abstract
Using wavelet packet decomposition, the energy coefficients in the fifth level of viral protein sequences were achieved to predict interspecies transmission. Since avian-origin influenza viruses could have high sequence similarities with human-origin avian influenza virus and could have the phenotype of interspecies transmission, viral data should be filtered to prevent the misconduct of feature selection and false performance of predicting models. Considering the balance of data size, the empirical cut-off value 97% was used to screen avian-origin influenza virus with high sequence similarity. The excellent performances of cross validation show that the SVM model has the best capability of predicting transmission and evaluating the contribution of five amino acid factors. The robust model was finally used to evaluate the filtered data of avian-origin virus and the results confirmed that double check for ambiguous phenotype of avian-origin virus with high sequence similarity was necessary and part of them have the ability to across species barriers.
KEYWORDS:
Avian influenza virus; Interspecies transmission; Machine learning; Wavelet package
PMID: 30528510 DOI: 10.1016/j.compbiolchem.2018.11.029