Tutorial

Web Search Interface

1. Select recipient organisms to include in the search.
   
   The "Simple Search" selection area can be sorted alphabetically by species name or lineage to assist in finding desired genomes. Alternatively, whole taxonomic groups can be selected in the "Advanced Search" section. In addtion to pre-selected groups, organisms whose lineages contain one or more specific terms can be selected by pasting the desired terms into the search text box. This feature can be used to choose sets of related or unrelated genomes.

2. Select LPI score range.
   
   Phylogenetically atypical proteins, which are the most likely potential candidates for horizontal gene transfer, can be detected by selecting only those matches with particularly low LPI scores (e.g. MAX < 0.6). Conversely, to find proteins that would be phylogenetically UNLIKELY as horizontal transfer candidates, a higher LPI score range can be selected (e.g. MIN > 0.75). LPI scores reflect phylogenetic distance of the match sequence from the query organism. Match organisms at similar phylogenetic distances recieve similar LPI scores, regardless of their database abundance. This feature is helpful in compensating for database bias in number of sequences associated with different taxonomic groups.

   As a guide to LPI score selection, LPI score frequencies for 955 microbial genomes are shown below, binned in increments of 0.05 score units. Proteins with LPI scores below 0.6 typically have no database matches closer than the phylum or class level, indicating strong phylogenetic discordance. LPI scores greater than 0.75 indicate that database matches exist in the same phylogenetic family, suggesting horizontal gene transfer is unlikely to be detectable by phylogenetic methods. Intermediate scores are typically borderline cases, which may be difficult to interpret. In some cases, available data may be insufficient to resolve whether or not HGT has actually occurred.




Optimal cutoff points may vary for individual genomes, depending on branch lengths of the phylogenetic trees underlying their lineage descriptions, as well as phylogenetic distance between available sequenced genomes. Users may wish to adjust LPI score cutoffs after viewing the genome-specific histogram presented on the DarkHorse results summary page for their organism of interest. Genome-specific LPI histograms usually contain several obvious break points in the distribution, as shown in the following examples (binned in score increments of 0.02 units).


LPI Score Histogram for Thermotoga lettingae




LPI Score Histogram for Thermoplasma acidophilum



3. Select phylogenetic granularity.
   
   Phylogenetic granularity describes the breadth of "self" sequences excluded as possible matches during a DarkHorse search. HGT events of different ages can be targeted by choosing different levels. Relative age of potential HGT events can be explored for a particular protein of interest by comparing its LPI scores at different phylogenetic granularities and looking for a point where the score changes from high to low.

   Strain level is the narrowest possible granularity choice. This setting is most useful for identifying relatively recent HGT events, because the low LPI proteins it finds are unique to a particular strain, not present in other strains of the same species. Using Burkholderia cenocepacia AU 1054 as an example, strain level granularity excludes all matches to database entries specifically labeled as belonging to strain AU1054, while still permitting matches to organisms like Burkholderia cenocepacia PC184 and Burkholderia cenocepacia MC0-3. Matches to these other strains of Burkholderia cenocepacia will receive high LPI values, but matches unique to strain AU1054 will receive lower LPI values.

   Species level granularity widens the list of excluded organisms. It can identify older HGT events by finding low LPI proteins unique to a particular species, absent from all other sequenced species of the same genus. In the previous example, this level of granularity would exclude matches to all strains of Burkholderia cenocepacia, but still allow matches to Burkholderia dolosa and Burkholderia xenovorans.

   Genus level granularity widens the exclusion list still further, excluding matches from all species within the same genus. In this case, no matches would be allowed to any members of the Burkholderia genus. Low LPI sequences found in only one genus, absent from all other known members of the same family, may represent more ancient HGT events.

   Users should keep in mind that predicting the age of HGT events requires some caution, because phylogenetically atypical proteins with low LPI scores can occur for two different reasons: either gene gain by the query organism, or gene loss from its closest known relatives. In cases where the query organism is the only sequenced example at a particular taxonomic level, it may not be possible determine age of gene acquisition until more data on related organisms becomes available.

4. Add additional filters (optional).

• Select protein functions or families by one or more annotation keywords (e.g "transporter").

• Select protein functions or families by sequence similarity to one or more reference proteins by amino acid BLAST search.

• Select only those matches where potential donor sequences are phylogenetically typical within their own taxa. This will reduce sensitivity, but provides more stringent selection against potential false positives. Excluding matches with reciprocal LPI scores < 0.75 typically eliminates HGT candidates that cannot be easily corroborated using phylogenetic trees.

• Select unusual GC content in coding sequence DNA by using z-score statistics, or by absolute percent GC. A z-score of 1.0 means that GC content was either higher or lower than the mean for all coding sequences in the genome by 1.0 standard deviations.

5. Select output format.

   Coding sequence location coordinates can indicate whether horizontal transfer candidates are adjacent or distant from each other, or located on the same chromosome, scaffold, or plasmid. To facilitate cross-referencing between protein and nucleic acid sequences, this option includes corresponding nucleic acid scaffold id numbers and coding sequence locus ids.

   DNA composition statistics include percent GC for each individual protein coding sequence, as well as mean and standard deviation for percent GC of all coding sequence regions in the parent genome. Percent GC is sometimes used as a simple marker of foreign DNA within a genome, although there may be a wide disparity between individual genes due to other factors.

   BLAST match statistics include alignment length, percent identity, e-value, and bitscore, as well as percent of the protein covered by the alignment. It also includes number of database matches included as potential orthologs during DarkHorse analysis. A high number of database matches indicates a protein that is conserved and well represented in database. Unusual or rapidly evolving proteins have fewer database matches. The "best" match sequence selected by the DarkHorse algorithm represents the closest database relative of potential donor organisms, whether by vertical or horizontal transmission.

6. View search results.

   For each individual genome, two types of output are available, a genome summary page and a tab-delimited file of raw, unfiltered results. The tab-delimited file can be downloaded and imported into a spreadsheet program such as Microsoft Excel. The summary page, in html format, includes a histogram of genome-wide LPI scores, a tally of match numbers by species, and statistics on total number of matched and unmatched proteins. It also includes phylogenetic lineage of the genome, as defined in the NCBI taxonomy database, and search-specific keywords and/or NCBI taxonomy ID numbers used by the DarkHorse program to eliminate self-matches to the query genome.

   Advanced results include composite data for all genomes selected, filtered according to parameters specified by the user. These results may be downloaded as a tab-delimited file, or viewed online in html format. Default result fields include information on both query and match sequences.

   Advanced results can be used to look for "reciprocal" LPI score relationships (low LPI query coupled with a high LPI match partner). Although LPI scores have only been calculated for matches belonging to publicly available, sequenced genomes, if a reciprocal LPI relationship can be detected it will show whether a potential donor sequence is typical or atypical within its own taxon. This information is useful in predicting whether or not sufficient data will be available to build a full-scale phylogenetic tree supporting horizontal gene transfer. Based on the LPI score distributions shown above, a combination of query LPI score less than 0.6 with a match partner (potential donor) LPI score greater than 0.75 is a reasonable starting place to select the best-supported HGT candidates.

References

1. Podell, S and Gaasterland, T (2007). DarkHorse: A method for genome-wide prediction of horizontal gene transfer. Genome Biology 8(2):R16
2. Podell, S Gaasterland, T, and Allen, EE (2008). A database of phylogentically atypical genes in archaeal and bacterial genomes, identified using the DarkHorse algorithm. BMC Bioinformatics 9:419