Introduction

Motifs are regions (or subsequences) of deoxyribonucleic acid (DNA) or protein sequences with a specific structure. These motifs generally contain functionally important sequences and are therefore often used as a base to classify proteins.

Two main approaches can be used to search and discover motifs: the enumerative approach and the probabilistic approach [1]. The MEME utilizes a probabilistic approach and is one of the most widely used software programs for the identification of novel “sequences” in sets of biological sequences [2–6]. The results generated by MEME command line software include figures of each motif and three other files, an .xml file, a .txt file, and an .html file. The .html file is a nice output for visual inspection of the results. However, the user cannot freely combine the figures they want from the .html file and directly generate them with MEME. The .txt file and .xml file contain almost all the information of the .html file. However, these files are difficult to read and cannot be used to directly generate figures and tables that meet the quality demands of publication standards. Some R/Bioconductor packages have been developed to process the results from MEME [7–9]. However, these R/Bioconductor R packages cannot parse the position of each motif on the corresponding sequences from the latest MEME. For example, the functions read_meme and importMeme from the Bioconductor packages universalmotif and memes, respectively, can only parse the information from the .txt file and cannot extract the position information. Their result is a list that is slightly unfriendly for the user. Additionally, the results produced by the MEME are difficult to combine with other downstream analysis software, such as ggtree [10]. The Bioconductor package motifstack can plot stacks with a hierarchical tree of the corresponding motifs [9]. However, in many cases, the user wants to visualize the phylogenetic tree with the corresponding position of each motif on the corresponding sequences.

Therefore, in this study, we developed the easy-to-use R package ggmotif to facilitate the extraction and visualization of results from the multiple file formats generated by the MEME and facilitate the integration of the data with other phylogenetic visualization tree packages, such as ggtree.

Materials and methods

In this study, we picked up the sequences of the AP2 gene family as example. The AP2 gene family, a large gene family of transcription factors, belongs to the AP2/ERF superfamily that plays important role during the lifespan of plant [11]. The sequences of the AP2 gene family of Arabidopsis thaliana were downloaded from PlantTFDB [12]. MEME (V5.4.1) was used to identify motifs with the parameters -protein -o meme_out -mod zoops -nmotifs 10 -minw 4 -maxw 7 -objfun classic -markov_order 0. Clustalo (V1.2.4) [13] and FastTree (V2.1.10) [14] were used to align the sequences and construct the phylogenetic tree. The bash code can be found at GitHub (https://github.com/lixiang117423/ggmotif).

The ggmotif workflow is shown in Fig 1. Briefly, the MEME can generate several files, including an .xml file and a .txt file containing the information of the location on sequence (s) and the sequence of motifs, respectively. The function getMotifFromMEME can extract the information described above and convert it to dataframe-type data objects that are user-friendly to biologists in the R environment. After that, the function motifLocation can be used to visualize the location of motif(s) on sequence(s), such as gene(s). If the user wants to plot the location of motif(s) on sequence(s) with the corresponding phylogenetic tree, the parameter tree will be used to meet the need with the help of ggtree. The MEME can make two type files, an .eps format figure and a .png format figure for motif sequence logo. That is enough for some users but not for others. After extracting the information of sequences of motifs, the user can filter or select motif(s) of interest to visualize with the function ggseqlogo from the ggseqlogo package [15].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. ggmotif workflow.

The ggmotif workflow consists of several functions, including the ggmotif package, ggtree package, and ggseqlogo package.

https://doi.org/10.1371/journal.pone.0276979.g001

Results

Visualization of motif location on sequences

After obtaining the motif information, the user can then visualize the sequence logos or the location of the motifs on the corresponding sequences. Conversely, the function motifLocation can be used to visualize the location of the motif(s) on the corresponding sequence(s) (Fig 2). The reference code is shown below:

1. library(ggmotif)
2. filepath <- system.file("examples", "meme.xml", package = "ggmotif")
3. motif_extract <- getMotifFromMEME(data = filepath, format = "xml")
4. motif_plot <- motifLocation(data = motif_extract)
5. motif_plot

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. The location of motifs on the corresponding sequences.

https://doi.org/10.1371/journal.pone.0276979.g002

Visualization of motif location on sequences with a phylogenetic tree

If the user has the corresponding phylogenetic tree, the id(s) is the same as the id(s) in the sequences used to identify motifs using MEME, the function motifLocation with parameter tree can be used to visualize the location of motif(s) on the corresponding sequence(s) after sorting the sequence(s) order with a phylogenetic tree. For example, as shown in Fig 3A, AT4G36920.2 is the id of the sequence used to identify motifs using MEME and construct the phylogenetic tree. So do other sequences. The figure generated by the parameter tree is the same as the above, but the order of sequences(s) is adjusted to fit with the phylogenetic tree (Fig 3A). The id of the phylogenetic tree requires that all formats supported by Bioconductor package ggtree must be consistent with the sequences inputted to MEME. If other information of phylogenetic tree is available, the parameter tree.anno can be used to plot the tip-point of phylogenetic tree (Fig 3B). As shown in demo result (Fig 3B), when the motifs positions are similar, the genetic relationship between the sequences is also similar. The reference code is shown below:

1. library(ggmotif)
2. filepath <- system.file("examples", "meme.xml", package = "ggmotif")
3. treepath <- system.file("examples", "ara.nwk", package = "ggmotif")
4. motif_extract <- getMotifFromMEME(data = filepath, format = "xml")
5. motif_plot <- motifLocation(data = motif_extract, tree = treepath)
6. motif_plot
7. tree.anno.path <- system.file("examples", "tree.anno.txt", package = "ggmotif")
8. tree.anno = data.table::fread(tree.anno.path) %>%
9.  dplyr::mutate(Group = as.character(Group))
10. motif_plot <- motifLocation(data = motif_extract, tree = treepath, tree.anno = tree.anno)
11. motif_plot +
12.  ggsci::scale_fill_aaas()

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3.

The location of motifs on the corresponding sequences (right panel) and the corresponding phylogenetic tree (left panel). The light blue and orange dots in Fig 3B represent the number of AP2/ERF domains in the corresponding sequences.

https://doi.org/10.1371/journal.pone.0276979.g003

Conclusion

The search and characterization of motifs for the given sequence(s) provides valuable insights into their role in biological regulation. Unlike previously developed tools, such as TBtools [16], ggmotif gives the user familiar with command line tool R a chance to extract and visualize motif information on the corresponding sequence(s). The Bioconductor package motifStack can plot motifs using tree-like structures, but it cannot visualize motif locations on corresponding sequences with phylogenetic trees. In other tools, it is difficult for the user to filter or select the motif of interest to visualize, and the user must almost handle the whole process if the motif of interest is changed. Overall, ggmotif is a user-friendly R package that provides functions to extract and visualize motif information from MEME.

Supporting information

Acknowledgments

We thank Dr. Jianming Zeng (University of Macau) and all the members of his bioinformatics team and biotrainee for generously sharing their experiences and codes. We would also like to thank TopEdit (www.topeditsci.com) and AJE (https://www.aje.com/go/plos/) for the English language editing of this manuscript.