Feature selection methods for identifying genetic determinants of host species in RNA viruses

Abstract

Despite environmental, social and ecological dependencies, emergence of zoonotic viruses in human populations is clearly also affected by genetic factors which determine cross-species transmission potential. RNA viruses pose an interesting case study given their mutation rates are orders of magnitude higher than any other pathogen--as reflected by the recent emergence of SARS and Influenza for example. Here, we show how feature selection techniques can be used to reliably classify viral sequences by host species, and to identify the crucial minority of host-specific sites in pathogen genomic data. The variability in alleles at those sites can be translated into prediction probabilities that a particular pathogen isolate is adapted to a given host. We illustrate the power of these methods by: 1) identifying the sites explaining SARS coronavirus differences between human, bat and palm civet samples; 2) showing how cross species jumps of rabies virus among bat populations can be readily identified; and 3) de novo identification of likely functional influenza host discriminant markers.

Figure 1. Feature selection of host specific genetic signatures within Flaviviridae.

The scatterplots display the first two principal components of the PCA undertaken using allele frequency information from (a) Flaviviruses' full polymerase sequences and (b) an alignment of the amino acids selected by the random forest algorithm. The maximum likelihood phylogenetic tree obtained from full polymerase sequences is presented in (c). Tree branch lengths reflect the number of amino acid differences per sequence.

Figure 2. SARS coronavirus species transitions and evolution.

The first two principal components of the PCA undertaken using (a) SARS coronavirus complete spike protein nucleotide sequences, and (b) nucleotides selected by the RFA. Viral groups, defined by host species and season, are represented by ellipses of different colours: Human patient samples from 2002/2003 collected in early, mid and late epidemic phase are HP03E (green), HP03M (purple) and HP03L (yellow); 2004 Human samples are labelled HP04 (black); palm civets samples collected in 2003 and 2004 are labelled PC03 (blue) and PC04 (red); bat samples are labelled BT (magenta).

Figure 3. Allele importance for host reservoir classification of SARS-like coronaviruses.

The alleles which were identified as significant for classification by the feature selection algorithm are represented by red points.

Figure 4. Cross-species transition events of rabies viruses in bats.

The first two principal components of the PCA undertaken using (a) complete Rabies virus nucleoprotein sequences, and (b) an alignment of nucleotides selected by the RFA. The ellipses of different colours represent the bat species in which virus samples were collected. Eight putative cross-species transmission events are highlighted in yellow with the respective predicted bat species of origin shown in (c).

Figure 5. Allele diversity across samples of influenza A H1N1 HA sequences collected from human (pre and post 2009 pandemic) and swine hosts.

Each vertical stripe represents allelic variance for a specific amino acid residue in three blocks of 40 sequences (taken at random) per host/virus type. The block of amino acids marked by an asterisk refers to the 100 residues to which the RFA has attributed the highest significance in explaining the allelic differences observed between groups. The ordering of other amino acids follow that of the HA gene. For each position (column) the allele present in the first human (seasonal) virus is colored blue. Moving from bottom to top, different alleles at the same position are then sequentially colored green, red, cyan, yellow and purple. Non polymorphic sites are not shown.

Figure 6. Computationally predicted structure of the 531–738 subset of amino acids in the PB2 subunit of the polymerase protein of influenza A viruses.

For structural prediction we used the consensus sequence for the subset of viruses' samples collected from each host species. These sequences contain two functional domains: the 627 (in cyan) and the NLS binding (in grey) domains. Highlighted in yellow are the amino acids which were identified by the RFA as discriminating host species.