Volume 8 Supplement 2
ProfileGrids: a sequence alignment visualization paradigm that avoids the limitations of Sequence Logos
© Roca; licensee BioMed Central Ltd. 2014
Published: 28 August 2014
The 2013 BioVis Contest provided an opportunity to evaluate different paradigms for visualizing protein multiple sequence alignments. Such data sets are becoming extremely large and thus taxing current visualization paradigms. Sequence Logos represent consensus sequences but have limitations for protein alignments. As an alternative, ProfileGrids are a new protein sequence alignment visualization paradigm that represents an alignment as a color-coded matrix of the residue frequency occurring at every homologous position in the aligned protein family.
The JProfileGrid software program was used to analyze the BioVis contest data sets to generate figures for comparison with the Sequence Logo reference images.
The ProfileGrid representation allows for the clear and effective analysis of protein multiple sequence alignments. This includes both a general overview of the conservation and diversity sequence patterns as well as the interactive ability to query the details of the protein residue distributions in the alignment. The JProfileGrid software is free and available from http://www.ProfileGrid.org.
Molecular biologists can learn about a protein's structure and function by studying the natural sequence variation resulting from a protein family's evolution . A multiple sequence alignment (MSA) is crucial for such bioinformatic analysis to relate homologous residues to one another as pioneered by early molecular evolution studies (for example ). Software tools are now available to align very large data sets consisting of 100,000+ sequences . However, there is a lack of programs that allow users to visualize and to interact with very large MSAs . Existing MSA visualization tools  continue to use a simple stacked sequence alignment visualization paradigm representing all individual sequences as rows in a table and homologous residue positions as columns. This "row-column" paradigm was sufficient decades ago when alignments were small; but, the stacked sequence representation is now inadequate for the amount of data being used today.
Sequence Logo Derivatives
Despite their widespread use, though, Sequence Logos have their limitations. While Logos are suitable for nucleic acids with only four residue symbols legible at even small sizes, there are problems when representing 20 residues of protein sequences. Variable regions of a protein alignment and positions with rare residues are not legible in Logos since the symbol size is small when scaled relative to the conserved positions. This has been described as a "totally incomprehensible jumble of letters."  However, even for conserved residues, the stacking of the symbols can lead to confusion. For example, Schneider pointed out that an "F" on top of an "L" could be mistaken for a long "E."  Logos do not display any information about residues missing from alignment columns and also lack a representation for gap symbols. Notably, Logos suffer from aesthetic challenges when two visualization channel types  are combined: symbol/stack height versus symbol color. Interpreting the tall/short stacked letter columns can be complicated by the color schemes used to distinguish different amino acid residue classes (such as the prominent red coloring of acidic protein residues). Finally, a recent user study evaluated the effectiveness of Sequence Logos for comparing motifs . The authors identified these problems: difficulty in judging the height of stacked symbols, no standardization for symbol color schemes, and a lack of interactivity for most current software tools.
The challenge in the visualization of large sequence alignments is in identifying a paradigm that summarizes the overall conservation trends while still providing easy navigation to detailed views of the underlying data . ProfileGrids as a MSA visualization paradigm were invented in 2005 for the analysis of the bacterial RecA protein family. MSA analysis had become impractical using the standard stacked sequence representation since a curated alignment had grown to several hundred homologs . A ProfileGrid reduces an alignment to a matrix, color-shaded according to the residue frequency in the MSA . The JProfileGrid Java program was upgraded to version 2.0  with new software features as well as improvements to the aesthetics in the visualization paradigm after receiving feedback from the http://VIZBI.org community . The ProfileGrid paradigm has two simple but significant differences that avoid Sequence Logos limitations: 1) all residue symbols are the same height since the matrix row sizes are fixed; 2) color shading of the cells is used to represent the frequency distribution of the residues in the sequence data. Thus, the overall conservation trends can be seen from the "heat map." The JProfileGrid software tool is an interactive MSA viewer taking advantage of the matrix representation of alignments.
The 2013 BioVis conference Redesign Contest provided an opportunity to demonstrate the usefulness of the ProfileGrid paradigm by visualizing the adenylate kinase lid (AKL) protein family alignments . Here, I report the final figures generated by the JProfileGrid software and the unique observations made possible by ProfileGrid analysis.
ProfileGrids clearly visualize protein residue distributions
The 2013 BioVis Redesign Contest description "recognizes the importance of effective encodings and clear visual communication in display of complex quantitative information" and "gives participants the opportunity to develop a practical replacement to the long-standing convention of sequence logos"  as exemplified by the AKL domain Sequence Logos (Figure 1). ProfileGrids are a practical and effective replacement for Sequence Logos when analyzing protein alignments. Furthermore, the mature JProfileGrid MSA software viewer has a rich graphical user interface allowing a molecular biologist to analyze their protein family of interest .
Overall trends of sequence conservation and variability are quite clear from the ProfileGrids and the user can choose whether to show the details of the residue frequencies (values in panel Figures 2B and 2C) or to show just the residues observed (Figure 2A) at that MSA position, i.e. each column. Note that the ProfileGrid in Figure 2A can be directly related to the Sequence Logo (Figure 1A) where the residues shown are stacked upon one another in each column; and, the symbol height reflects the conservation of each residue. The ProfileGrid, however, allows each symbol to be legible which is critical for interpreting a variable column such as position 14. The ProfileGrid allows the entire protein mutation distribution to be assessed. Similarly, rare residues are obvious such as at position 8 (Figure 2B) where the Redesign Contest instructions incorrectly stated that "glycine is absolutely conserved in the Gram-negatives."  In fact, 2 sequences have asparagine at that position (Theileria parva strain Muguga [GenBank:XP_766154.1 http://www.ncbi.nlm.nih.gov/protein/71033025] and Theileria annulata [Genbank:XP_954152.1 http://www.ncbi.nlm.nih.gov/protein/XP_954152.1]). Such rare observations would be impossible to identify from the Sequence Logo paradigm (Figure 1B) or its derivatives (the exception being InverseLogos depending upon user settings ). An AKL protein expert can now interpret whether the Gram negative homologs containing 8-asn are interesting natural sequence variation exceptions sharing a residue with 18 Gram positive homologs (Figure 2C) or are just sequence errors resulting from experimental problems during data collection ). The interactive JProfileGrid user interface made it trivial to identify the two asparagine-containing sequences since the user can select any cell in the ProfileGrid window and then perform a query of the MSA. Thus, the ProfileGrid paradigm enables the MSA to be treated as a searchable database of sequence records.
ProfileGrids can effectively visualize alignment differences
The 2013 BioVis Redesign Contest challenged applicants to help the AKL biologists understand the differences between the Gram negative (Figure 1B) and Gram positive (Figure 1C) sequences by creating alternatives to the Sequence Logo representation. Visualizing the protein sequence differences between the two subpopulations would presumably allow the biologists to propose structure and function hypotheses about the AKL protein activities. The original figure designed by the AKL biologists is of three stacked Sequence Logos (Figure 1). Drawing comparisons between protein Sequence Logos is very difficult especially since there is no visual encoding for differences (although attempts have been made to introduce such a feature [11, 17, 18]). In particular, a user will be frustrated performing mental comparisons on adjacent Sequence Logos such as when comparing the Gram negative (Figure 1B) and Gram positive (Figure 1C) sequences. Such challenges were documented in a recent Sequence Logo user study . By contrast, the interactive JProfileGrid software is especially suited for comparing sequences populations to one another.
The ProfileGrid paradigm can allow color coding to visually represent amino acid classes as used in Sequence Logos. However, in this case the coloring is minimized to provide just enough information rather than dominate the figure and confuse the user. Figure 3A shows a color scheme to represent hydrophobic (brown), basic (blue), nitrogen-containing (lavender), acidic (red), and other (green) residues. Note that brown is used instead of black (as in the default AKL Sequence Logo coloring of Figure 1) to avoid visually clashing with the frequency grayscale encoding. Thus, as a reader scans across the ProfileGrid, the row position will encode the amino acid class in a manner complementary to the frequency information.
The ProfileGrid highlight (pink cell border) feature takes advantage of "visual popout"  to allow the user to make observations. AKL alignment position 9 is a good case study to examine differences between the Gram negative and Gram positive subpopulations. In Figure 3B arginine (positively charged coded "blue," middle rows as per Figure 3A) is very well conserved (black cell; row 13, column labelled "9") in Gram negatives. However, an alanine (Figure 3B; pink borders; row 3, column labelled "9") is found in Gram positives and this residue is small and hydrophobic (brown, top rows). Intriguingly, examination of the Gram positive sequences (Figure 3C) shows that there are at least two predominant subpopulations observed (see Figure 2C for the complete residue distribution at position 9): the aforementioned alanine (gray box; row 3, column 9) but also the arginine (gray box with pink borders; row 13, column 9) shared with the Gram negatives. Perhaps the AKL experts can determine if there is any biological significance to these qualitative observations; but, this shows that the ProfileGrid paradigm allows these observations to be made at all.
JProfileGrid program features
The interactive JProfileGrid program viewer has features designed for the biologist user [30, 31]. In software version 2.0, a new "overview" mode allows the visualization of the entire MSA within one window as either a ProfileGrid or as stacked sequences. Individual ProfileGrid cells can be selected to extract sequence subsets of interest during a visualization dissection. Sorting the residue rows by physical/chemical properties such as flexibility, helix propensity, hydropathy, and volume allow qualitative structural analyses to be performed. The detailed ProfileGrid window with the symbol counts, has a new second pane to view different parts of the MSA at the same time. The "highlight" feature can identify residues that occur greater or less than a user-defined threshold of residue frequency. Large alignments can be separated into subsets of interest by using metadata filtering once JProfileGrid imports simple sequence annotations from flat file spreadsheet databases. The interactive features of the JProfileGrid program can be more easily appreciated by a live walk-through as demonstrated by my 2013 BioVis Data Contest movie submission .
ProfileGrid advantages over Sequence Logos/Bundles
ProfileGrids avoid the limitations of Sequence Logos especially for protein MSAs. Namely, all ProfileGrid residue symbols are legible so that no information is lost for bioinformatic analysis. ProfileGrids display gap symbols as well as the regions of a MSA where residues are not observed. Neither case is handled by Sequence Logos. The ProfileGrid paradigm, in the opinion of this author, clearly and effectively separates visual encoding channels to be more aesthetically pleasing than Sequence Logos. Such clarity allows for the careful dissection of sequence conservation patterns in MSAs by molecular biologists. Importantly, ProfileGrids solve the visualization problem of handling very large alignments since there is almost no limit to the number of sequences that can be represented. The matrix representing the MSA is only 21 rows rather than the inefficient stacked sequence representation that lists every individual sequence as a row. While a Sequence Logo can also represent very large MSAs, the details of the underlying sequence information are lost as in most consensus paradigms. The interactive JProfileGrid viewer allows the user to retain access to the sequence data for a robust protein family analysis.
Finally, we comment on the Redesign Contest applicants. Most of the submissions were derivatives of Sequence Logos and so would have the same aforementioned limitations. Interestingly, the two entries that received Honorable Mentions [35, 36] have replaced the Sequence Logo representation with a different visualization paradigm that one entrant named "Sequence Bundles." We note that by representing the MSA as a matrix with residue symbols as rows and the alignment positions as columns, they are handling the underlying data as a profile for bioinformatic sequence similarity searches  using command-line programs with no graphical user interfaces. The 2008 ProfileGrid publication, to the best of my knowledge, is the first demonstration of profiles as a visual paradigm for analyzing MSAs. How Sequence Bundles differ significantly from ProfileGrids is in the former's use of lines to connect matrix cells along the path of an individual protein sequence. More sequences occurring in an alignment result in lines stacking on the matrix visualization allowing consensus sequences to emerge as a thick bundle.
While the Sequence Bundle connecting lines are aesthetically interesting, they are also a liability since as the MSAs get very large, the visualization will become more cluttered. In my opinion, what is relevant to the biologist user within the context of an MSA is the frequency of the residues in the visualization. After an observation has been made, then the user will dig deeper to identify the particular sequences of interest. The interactive JProfileGrid program was designed for such a task. Individual sequences are shown alongside the top row of the ProfileGrid by choosing the desired reference and highlight sequences from JProfileGrid menus. Selecting individual ProfileGrid cells allow a separate JProfileGrid window to display the subset of the sequences containing that residue in the MSA. There is no need for connecting lines to be layered on top of the matrix as in Sequence Bundles.
While Sequence Logos have their merits for visualizing short conserved motifs especially in nucleic acid sequences, this paradigm has limitations when representing protein alignments. The ProfileGrid paradigm replaces Sequence Logos and solves the challenge of visualizing large protein alignments. Thus, ProfileGrids allow a molecular biologist to clearly and effectively analyze protein structure and function.
List of abbreviations used
adenylate kinase lid
multiple sequence alignment
This study was self-financed by AIR. I gratefully acknowledge the dataset provided by Drs. Magliery and Sullivan at The Ohio State University for the purposes of the BioVis 2013 Contest. I also thank the 2013 BioVis contest organizers especially William Ray, Ph.D., of The Ohio State University.
Publication of this work was self-financed by AIR.
This article has been published as part of BMC Proceedings Volume 8 Supplement 2, 2014: Proceedings of the 3rd Annual Symposium on Biological Data Visualization: Data Analysis and Redesign Contests. The full contents of the supplement are available online at http://www.biomedcentral.com/bmcproc/supplements/8/S2
- Higgins RF: Of URFs and ORFs: A Primer on How to Analyze Derived Amino Acid Sequences. 1986, University Science BooksGoogle Scholar
- Fitch WM: An improved method of testing for evolutionary homology. J Mol Biol. 1966, 16: 9-16.View ArticlePubMedGoogle Scholar
- Sievers F, Wilm A, Dineen D, Gibson TJ, Karplus K, Li W, Lopez R, McWilliam H, Remmert M, Soding J, Thompson JD, Higgins DG: Fast, scalable generation of high-quality protein multiple sequence alignments using Clustal Omega. Molecular Systems Biology. 2011, 7: 539-View ArticlePubMedPubMed CentralGoogle Scholar
- Procter JB, Thompson J, Letunic I, Creevey C, Jossinet F, Barton GJ: Visualization of multiple alignments, phylogenies and gene family evolution. Nature Methods. 2010, 7: S16-25.View ArticlePubMedGoogle Scholar
- List of alignment visualization software. [http://en.wikipedia.org/wiki/List_of_alignment_visualization_software]
- Schneider TD, Stephens RM: Sequence logos: a new way to display consensus sequences. Nucleic Acids Research. 1990, 18: 6097-6100.View ArticlePubMedPubMed CentralGoogle Scholar
- Gorodkin J, Heyer LJ, Brunak S, Stormo GD: Displaying the information contents of structural RNA alignments: the structure logos. Comput Appl Biosci. 1997, 583-6. 13
- Schuster-Böckler B, Schultz J, Rahmann S: HMM Logos for visualization of protein families. BMC Bioinformatics. 2004, 5: 7-View ArticlePubMedPubMed CentralGoogle Scholar
- Fujii K, Zhu G, Liu Y, Hallam J, Chen L, Herrero J, Shaw S: Kinase peptide specificity: improved determination and relevance to protein phosphorylation. Proc Natl Acad Sci USA. 2004, 101: 13744-13749.View ArticlePubMedPubMed CentralGoogle Scholar
- Workman CT, Yin Y, Corcoran DL, Ideker T, Stormo GD, Benos PV: enoLOGOS: a versatile web tool for energy normalized sequence logos. Nucleic Acids Res. 2005, W389-92. 33
- Vacic V, Iakoucheva LM, Radivojac P: Two Sample Logo: a graphical representation of the differences between two sets of sequence alignments. Bioinformatics. 2006, 22: 1536-7.View ArticlePubMedGoogle Scholar
- Li W, Yang B, Liang S, Wang Y, Whiteley C, Cao Y, Wang X: BLogo: a tool for visualization of bias in biological sequences. Bioinformatics. 2008, 24: 2254-5.View ArticlePubMedGoogle Scholar
- Colaert N, Helsens K, Martens L, Vandekerckhove J, Gevaert K: Improved visualization of protein consensus sequences by iceLogo. Nat Methods. 2009, 6: 786-7.View ArticlePubMedGoogle Scholar
- Leipzig J: Making a better sequence logo. [http://chopcbmi.org/2011/11/10/making-a-better-sequence-logo/]
- Thomsen MC, Nielsen M: Seq2Logo: a method for construction and visualization of amino acid binding motifs and sequence profiles including sequence weighting, pseudo counts and two-sided representation of amino acid enrichment and depletion. Nucleic Acids Res. 2012, W281-7. 40
- O'Shea JP, Chou MF, Quader SA, Ryan JK, Church GM, Schwartz D: pLogo: a probabilistic approach to visualizing sequence motifs. Nat Methods. 2013, 10: 1211-2.View ArticlePubMedGoogle Scholar
- Beitz E: Subfamily logos: visualization of sequence deviations at alignment positions with high information content. BMC Bioinformatics. 2006, 21: 313-View ArticleGoogle Scholar
- Shih AC, Lee DT, Peng CL, Wu YW: Phylo-mLogo: an interactive and hierarchical multiple-logo visualization tool for alignment of many sequences. BMC Bioinformatics. 2007, 8: 63-View ArticlePubMedPubMed CentralGoogle Scholar
- Pérez-Bercoff A, Koch J, Bürglin TR: LogoBar: bar graph visualization of protein logos with gaps. Bioinformatics. 2006, 22: 112-4.View ArticlePubMedGoogle Scholar
- Bindewald E, Schneider TD, Shapiro BA: CorreLogo: an online server for 3D sequence logos of RNA and DNA alignments. Nucleic Acids Res. 2006, W405-11. 34
- Freyhult E, Moulton V, Ardell DH: Visualizing bacterial tRNA identity determinants and antideterminants using function logos and inverse function logos. Nucleic Acids Res. 2006, 905-16. 34
- Chang TH, Horng JT, Huang HD: RNALogo: a new approach to display structural RNA alignment. Nucleic Acids Res. 2008, W91-6. 36
- Menzel P, Seemann SE, Gorodkin J: RILogo: visualizing RNA-RNA interactions. Bioinformatics. 2012, 28: 2523-6.View ArticlePubMedGoogle Scholar
- Karwath A, Kersting K: Relational Sequence Alignments and Logos. Inductive Logic Programming. Edited by: Muggleton S, Otero R, Tamaddoni-Nezhad A. 2007, Springer, 290-304. [Hutchison D (Series Editor) Lecture Notes in Computer Science, vol 4455.]View ArticleGoogle Scholar
- Sharma V, Murphy DP, Provan G, Baranov PV: CodonLogo: a sequence logo-based viewer for codon patterns. Bioinformatics. 2012, 28: 1935-6.View ArticlePubMedPubMed CentralGoogle Scholar
- Crooks GE, Hon G, Chandonia JM, Brenner SE: WebLogo: a sequence logo generator. Genome Res. 2004, 1188-90. 14
- Munzner T: Visualization. Fundamentals of Computer Graphics. Edited by: Peter Shirley and Steve Marschner. 2009, AK Peters, 675-707.Google Scholar
- Lee NK, Oon Y: Potential perils of biological sequence visualization using Sequence Logo. Computer Graphics, Imaging and Visualization (CGIV) International Conference 6-8 August 2013. Edited by: Macau. Banissi E, Sarfraz M. 2013, IEEE, 106-111.Google Scholar
- Roca AI, Cox MM: RecA protein: structure, function, and role in recombinational DNA repair. Progress in Nucleic Acid Research and Molecular Biology. 1997, 56: 129-223. [http://www.sciencedirect.com/science/article/pii/S0079660308610053]View ArticlePubMedGoogle Scholar
- Roca AI, Almada AE, Abajian AC: ProfileGrids as a new visual representation of large multiple sequence alignments: a case study of the RecA protein family. BMC Bioinformatics. 2008, 9: 554-View ArticlePubMedPubMed CentralGoogle Scholar
- Roca AI, Abajian AC, and Vigerust DJ: ProfileGrids solve the large alignment visualization problem: influenza hemagglutinin example [v1]. F1000Research. 2013, 2: 2-[http://f1000research.com/articles/2-2/]Google Scholar
- BioVis 2013 Redesign Contest: Protein Mutation Profiles. [http://www.biovis.net/year/2013/info/redesign-contest]
- BioVis 2013 Redesign Contest "instructions.pdf". [http://www.biovis.net/sites/default/files/biovis2013.redesign.tgz]
- JProfileGrid Facilitates Visual Exploration of dTIM Mutations within the Context of a Family Alignment. [http://dx.doi.org/10.6084/m9.figshare.902160]
- Sakai R, Aerts J: Sequence diversity diagram for comparative analysis of multiple sequence alignments. BMC Proceedings. 2014, 8 (Suppl 2): S9-View ArticlePubMedPubMed CentralGoogle Scholar
- Kultys M, Nicholas L, Schwarz R, Goldman N, King J: Sequence Bundles: a novel method for visualising, discovering and exploring sequence motifs. BMC Proceedings. 2014, 8 (Suppl 2): S8-View ArticlePubMedPubMed CentralGoogle Scholar
- Gribskov M, McLachlan AD, Eisenberg D: Profile analysis: detection of distantly related proteins. Proc Natl Acad Sci USA. 1987, 84: 4355-8.View ArticlePubMedPubMed CentralGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.