RankVISTA

Fig. 1. A sample rankVISTA graph.

Fig. 2. A sample of rankVISTA regions.

RankVISTA conservation plots depict evolutionarily conserved segments in pairwise or multiple alignments as a bar graph, where the heights scale with statistical significance [-log₁₀(P-value)]. For example, a height of 4 indicates that the probability of seeing that level of conservation by chance in a neutrally-evolving 10-kb segment of the base sequence is less than 10^-4.

RankVISTA graphs are based on the Gumby algorithm, which estimates neutral evolutionary rates from non-exonic regions in the multiple sequence alignment, and then identifies local segments of any length in the alignment that evolve more slowly than the background. The phylogenetically weighted log-odds conservation scores of conserved segments are translated into P-values using karlin-altschul statistics. Gumby has no window-size parameter, and no fixed percent-identity threshold either. Since the algorithm uses a more-conserved-than background paradigm, it can perform phylogenetic shadowing (close species) and footprinting (distant species) with equal facility.

NOTE: Since the input alignment is its own training set, small or grossly incomplete alignments are to be avoided:

• The base sequence length should be at least 10 kb. Smaller alignments might be tolerated, but if Gumby detects an inadequate number of aligned positions, it will return no output.

• For the p-values to be meaningful, Gumby requires a reasonably complete alignment. Rule of thumb: the number of "N" characters and spurious gap characters arising from missing sequence data should be less than 10% of the total number of characters in the alignment. The relative ranking of conserved regions by pvalue would still be meaningful if this rule were violated. However, the p-value estimates would be systematically biased.

• Gumby's sensitivity in detecting non-exonic conservation can be increased by supplying exon annotations. The annotated regions are masked when estimating neutral evolutionary rates, resulting in a more accurate estimate of the background conservation level.

RankVista regions are colored according to their annotation, as seen in the figure above. Note that RankVista coloring is based on exon annotations of all aligned sequences, not just the one currently used as the base. Consequently, an unannotated region in the base sequence might still be colored as an exon because of annotations from other sequences. In another deviation from the standard scheme, RankVISTA colors UTRs and coding exons the same, since they are treated identically by Gumby.

References:

Martin J, et al. The sequence and analysis of duplication-rich human chromosome 16. Nature. 2004 Dec 23;432(7020):988-94.