Figure 1.
An example to illustrate the concepts of a mosaic and its binary encoding upon which the genetic algorithm operates.
Panel A: a phylogenetic breakpoint/lineage model which “threads” a query sequence (labeled ‘Q’) onto the reference tree with sequences. Panel B: the example individual model (mosaic)
is encoded by a 36-bit binary vector on 5 fragments (genes)–2 for placing the breakpoints (Gray-binary encoded) and 3 for identifying sister lineages, binary encoded using the post-order traversal scheme shown in the reference tree of Panel A.
Figure 2.
Algorithmic flowchart of SCUEAL.
Algorithmic logic underlying SCUEAL; see Figure 3 for a description of the genetic algorithm itself. Refer to the text for more detailed descriptions of individual procedures and parameter definitions.
Figure 3.
Algorithmic flowchart of the genetic algorithm in SCUEAL.
A flowchart description of the genetic algorithm applied to a given starting population and controlled by input parameter values. Refer to the text and Figure 2 for further description of individual steps and parameter definitions.
Figure 4.
A simulation scenario example.
One of the simulation scenarios used to asses our detection method with the results over replicates (scenario 5/close in Table 2). The query sequence (2) was simulated to move from reference lineage 1 to reference lineage 3 every 400 bp as shown in the tree panel. The clustering chart depicts model and replicate averaged support for assigning the query sequence to a particular reference lineage, as estimated by the genetic algorithm over 100 simulated data replicates, whereas black impulse plots indicate the inferred placements of breakpoints. The y-axis does not reach
because each replicate contributes the model averaged support for the best inferred mosaic type–a value that is
; the upper limit on the y-axis is, therefore, the mean (over replicates) model-averaged support for the best-fitting mosaic (0.92 in this case).
Table 1.
SCUEAL performance on simulated data.
Figure 5.
Power and accuracy in the sequence shuffling simulation.
Power of SCUEAL to detect breakpoints in the HIV-1 pol sequence shuffling scenario as a function of recombinant fragment length (x-axis) and divergence between parental strains (y-axis). Grid cells are colored according to the proportion of correctly detected breakpoints (different cells may summarize different numbers of simulations). White squares are plotted when there were no simulated breakpoints within a corresponding length-divergence range of values.
Figure 6.
An example of a good agreement between SCUEAL and REGA in classifying a partial pol subtype B sequence.
The SCUEAL clustering plots present in this figure and Figures 7, 8 and 9 are conceptually analogous to bootscan plots, i.e. which reference sequence is the most likely sister lineage of the query sequence for a given site, but is based on model averaged support values instead of phylogenetic bootstrap. A partial reference tree with placed query is shown; color coding is consistent between the similarity plot and the tree. A phylogenetic tree with bootstrap support values and bootscan plot using the REGA alignment generated for the query sequence are shown.
Figure 7.
An instance when a sequence unclassified by REGA is inferred to be a novel recombinant form by SCUEAL; the A–J mosaic structure is also confirmed by trees and bootscan plots based on the REGA reference alignment.
Figure 8.
An example of within-subtype (B) recombination detected by SCUEAL, but not by REGA. A partial reference tree with placed query is shown; color coding is consistent between the similarity plot and the tree.
Figure 9.
An instance when a sequence assigned to subtype A by REGA is deduced to be an A-B-A mosaic by SCUEAL.
Similarity plots based on the reduced REGA alignments (only A and B subtype reference sequences) confirm that the same mosaic structure is supported using if a small enough window is selected for a sliding window analysis.
Table 2.
SCUEAL screening results on partial HIV-1 reverse transcriptase sequences from the Stanford Drug Resistance database.
Table 3.
SCUEAL screening results on partial HIV-1 polymerase sequences from the UK.