Visualizing bacterial tRNA identity determinants and antideterminants using function logos and inverse function logos

Abstract

Sequence logos are stacked bar graphs that generalize the notion of consensus sequence. They employ entropy statistics very effectively to display variation in a structural alignment of sequences of a common function, while emphasizing its over-represented features. Yet sequence logos cannot display features that distinguish functional subclasses within a structurally related superfamily nor do they display under-represented features. We introduce two extensions to address these needs: function logos and inverse logos. Function logos display subfunctions that are over-represented among sequences carrying a specific feature. Inverse logos generalize both sequence logos and function logos by displaying under-represented, rather than over-represented, features or functions in structural alignments. To make inverse logos, a compositional inverse is applied to the feature or function frequency distributions before logo construction, where a compositional inverse is a mathematical transform that makes common features or functions rare and vice versa. We applied these methods to a database of structurally aligned bacterial tDNAs to create highly condensed, birds-eye views of potentially all so-called identity determinants and antideterminants that confer specific amino acid charging or initiator function on tRNAs in bacteria. We recovered both known and a few potentially novel identity elements. Function logos and inverse logos are useful tools for exploratory bioinformatic analysis of structure-function relationships in sequence families and superfamilies.

Figure 1

Function logos of potential identity determinants among bacterial tDNAs in the structurally aligned Modified Sprinzl Database (12). There is one function logo for each of the five possible states in an RNA alignment; Thymine is represented as Uracil. Letters show tRNA classes associated with the given sequence state at the respective position given by the x-axis. Annotating the x-axis in addition to position in our structural alignment are the locations of the standard stems and the Sprinzl position indexing (13). See text for more detail.

Figure 2

A geometric representation of how the simplex and reciprocal transforms map the different compositions of Table 1, which contain three components: A, B and C. In the 2-simplex with p(A) = 1 in the upper corner, p(B) = 1 in the lower left corner and p(C) = 1 in the lower right corner, the original data points are in black at the bottom of the figure and the transformed compositions are shown in color and labeled.

Figure 3

A comparison of the performance of the simplex and reciprocal transforms in generating inverse logos (center and right) from the hypothetical data listed in Table 2 and visualized in the original sequence logo shown on the left.

Figure 4

Inverse function logos of potential identity antideterminants among bacterial tDNAs in the structurally aligned Modified Sprinzl Database (12). There is one inverse function logo for each of the five possible states in an RNA alignment; Thymine is represented as Uracil. Letters show tRNA classes under-represented among sequences with the given sequence state at the respective position given by the x-axis. Annotating the x-axis in addition to position in our structural alignment are the locations of the standard stems and the Sprinzl position indexing (13). See text for more detail.