Revealing Molecular Targets for Enterovirus Type 71 Detection by Profile Hidden Markov Models

The enterovirus infection in 1998 claimed 78 deaths in Taiwan, with an average of 40 fatalities each year after. Traditional serum-based diagnostic methods often fail to detect enteroviruses due to antigenic changes. As a result, many isolates remain untyped and are absent from the enterovirus surveillance and epidemiological investigations. We present a profile hidden Markov model (HMM) method for molecular typing of enterovirus 71 (EV71). Based on the enteroviral sequences retrieved from GenBank, we build a nucleotide-based and an amino acid-based profile HMM for each EV71 gene using the package HMMER. HMMER bit score-based Z-scores for EV71 and non-EV71 sequences are calculated for each of these profile HMMs. In a genome-wide analysis, we find that the distribution of the EV71 Z-scores and that of the non-EV71 Z-scores have disjoint support for nucleotide-based VP1 profile HMM if the sequence is longer than 150 bases; a VP1-based molecular typing method for EV71 is thus proposed. We also report VP4 an alternative molecular target for detecting EV71, while the two UTRs and all the genes coding the internal proteins cannot be used for such purpose. To demonstrate the performance of the nucleotide-based EV71 VP1 profile HMM, 330 enterovirus VP1 nucleotide sequences newly reported to GenBank are typed with this method. All the EV71 sequences are detected with no error.