2.2. Generating Alignments

• Access Mulan via the internet at http://mulan.dcode.org/ (Fig. 1A). Alternatively access Mulan via the ECR.

  

  Open in a separate window

  Fig. 1

  Accessing the Mulan tool from the homepage at http://mulan.dcode.org (A) or from the ECR Browser “Synteny/Alignments” link (B).

  Browser at http://ecrbrowser.dcode.org, through the “Synteny/Alignments” link. Click on each box next to the sequence to be aligned and then click on the “Mulan” button provided at the bottom of the page (Fig. 1B).

• At the Mulan homepage, indicate the number of species that will be used in the analysis, and select the desired alignment type: tba-based (left button) or refine-based (right button) (Fig. 1A). (It is advised to select the tba-based approach if the user is unsure of which option is best suited, or have sequences in single-contig format. The tba-method includes more options and provides more sensitive alignments than refine.)

• Sequence input.

  1. Submit sequences in FASTA format and gene annotation in format described in the Mulan documentation, and select the appropriate option for masking repetitive elements (Fig. 2A). Although Mulan is capable of running Repeat-Masker locally (http://www.repeatmasker.org/) to mask repetitive elements in input sequences, submitting premasked sequences will significantly reduce the total processing time.

     

     Open in a separate window

     Fig. 2

     Each sequence can be pasted in, in FASTA format, uploaded as a FASTA file, or entered as an accession number along with the available annotation (A). Alternatively, sequences can be fetched from the UCSC Genome Browser individually using the “Upload” function (A), or in groups (Batch Upload System) Browser (B). Once sequences have been uploaded, the program acknowledges the receipt (C).

  2. If sequences of interest are available from fully sequenced genomes, Mulan can automatically fetch these sequences from the University of California Santa Cruz (UCSC) Genome Browser (http://genome.ucsc.edu/). To do so, the user needs to click “Upload” (Fig. 2A), and provide the necessary information for the automated upload feature to fetch the sequences directly from the UCSC Genome Browser (Fig. 2B). The required information includes: (1) the organism or genome to be used, (2) the assembly version, (3) the type of annotation, and (4) genome coordinates. Once Mulan downloads the sequence along with its annotation onto the server, the successful upload is acknowledged (Fig. 2C), and the alignment engine proceeds to create alignments between the input sequences. Note that the upload feature automatically extracts information on repetitive elements along with the sequence data; it also permits selection of different gene annotation sources. This automated upload can be combined with manual input of sequences missing representation in fully sequenced genomes.

  3. Alternatively, if all the sequences of interest are available from fully sequenced genomes, the “Batch Upload System or BUS” integrated into Mulan can be used to simultaneously fetch all the sequences at once. Follow the BUS link on top of the sequence upload page to access this feature (Fig. 2A).

• Step-by-step specifics for the tba-based alignment approach. Upon generating a set of preliminary pairwise alignments, a phylogenetic tree is presented to the user, who has the option to accept it by clicking the “Continue” button, or refine it, if it is believed that the tree does not accurately depict the relationship between the input sequences (Fig. 3). This phylogenetic tree will be guiding the construction of the full tba-based multiple-sequence alignment.

  

  Open in a separate window

  Fig. 3

  Mulan defines a guiding phylogenetic tree before proceeding with the detailed sequence alignment. The user has the option to submit modifications to this tree.

• Once the alignment request is completed, Mulan presents results data analysis on an interactive summary page (Fig. 4). The summary page consists of multiple links to the dynamic conservation profile visualization module, textual multiple sequence alignment (with a dynamically modified base sequence; specific to tba-based alignments), hot-link to multiTF detection of evolutionary conserved TFBS (specific to tba-based alignments), dot-plots describing sequence rearrangements, interactive selection of ECRs, etc. One also has the option to adjust annotation files and sequence titles from this page.

  

  Open in a separate window

  Fig. 4

  A completed alignment request results in a “summary page” that provides links to the interactive visualization platform, pairwise dynamic plots, dot plots, annotation files, sequence files, and a portal to the transcription factor binding site analysis software, MultiTF.

• The processing information is stored on our servers for a limited amount of time (usually up to 3 months) and the data can be reaccessed anytime from the homepage (Fig. 1A) by providing the job identification number (request ID) listed on the top left corner of the summary page.

2.3. Visualization and Data Analysis Strategies for Multisequence Local Alignments

Multiple-sequence comparative analysis is a challenging task in terms of generating highly reliable alignments and graphically displaying the alignment results. To address the complexity stemming from user input sequence files that potentially consist of a large number of sequences of varying lengths and different phylogenetic relationships, we provide a set of different visualization options in Mulan. In general, Mulan alignment visualization is based on the zPicture display design (6), where the reference sequence is linear along the horizontal axis and the percent identity is plotted along the vertical axis. All the dynamic visualization options can be accessed through the summary page (Fig. 4). The “Dynamic Visualization” link directs the user to the interactive alignment display (Fig. 5). At this page the top bar (Fig. 5A) allows the user to customize the visual display by selecting the desired:

1. The Graphical type of evolutionary conservation profile (smooth or percent identity plot).

2. The length of the sequence to be displayed per each line.

3. The size and percent identity of the ECR to be highlighted in the graphical alignment display.

4. The percent identity for the bottom cut-off.

5. The subregion to be indicated as “from” – “to” coordinates.

Open in a separate window

Fig. 5

Mulan interactive alignment customization options (A) and graphical display of alignments (B).

To assist in the visual analysis of conservation, Mulan has several additional options available.

1. The user can choose to color code ECRs that are present in a particular number of species (Fig. 5). This option will dynamically prioritize regions with variable degree of phylogenetic occurrence (see Note 2).

2. The user has the option to change the base genome in the visualization of multi-species sequence evolution. This provides the option to study conservation of regions specific to different lineages and closely related groups of species. By changing the base species, the new stacking order of conservation profiles with the rest of the species will be automatically determined using the evolutionary relationship of each sequence to the reference sequence, where more closely related species are at the bottom. (Option specific to the tba-based alignment.)

3. Visualization scheme provides the means to include or remove the legend in the display as well as to adjust the graph height.

4. Contig names and alignment blocks can be visualized as tracks on top of the conservation profile (Fig. 6). In this situation, syntenic blocks are color-coded based on their orientation in respect to the base sequence thus allowing for easy ordering-and-orienting of draft sequences by using the base sequence as the architectural guide. This feature can be used as a preassembly tool when multiple overlapping contigs are available from a homologous interval in a new species with detectable sequence similarity to the base sequence. (Option specific to the refine-based alignment.)

   

   Open in a separate window

   Fig. 6

   Contig ordering based on homology to the reference sequence. The top layer of shaded lines indicates the location of contigs from a second sequence aligned to the base sequence where right-turned triangles specify forward strand alignments, and left-turned triangles correspond to reverse strand alignments.

5. “Color density by interspecies conservation” illustrates the relationship between a conserved element and the number of species that share a particular region (Fig. 7A) such that, the more species share a sequence, the darker the conservation profile will be displayed. (This analysis is performed for every pixel-wide region of the conservation plot. The number of ECRs from different species that overlap with a particular pixel count toward the number of species sharing this region.)

   

   Open in a separate window

   Fig. 7

   Mulan alignment analysis options: color density by interspecies conservation (A) and summary of conservation display (B).

6. Similar to Picture, Mulan allows interactive and customized ECR analysis. Users can select the evolutionary criteria (length and percent identity) for graphical identification of ECRs from the conservation plot. We have previously shown that longer and well conserved ECRs can be indicators of functional elements in genomic alignments (14) and this option permits the user to prioritize and define the optimal amount of ECRs in the studied locus—to adjust for highly conserved vs poorly conserved loci.

7. Two additional data representation modules are implemented in the Mulan tool: phylogenetic shadowing and summary of conservation. Summary of conservation collects shared similarities from all the pairwise comparisons into a single conservation profile (Fig. 7B), the phylogenetic shadowing option effectively collects pairwise mismatches (11). Thus, the summary of conservation option will aid in reconstructing conservation profiles in cases of highly diverged sequences, whereas the phylogenetic shadowing option will facilitate the identification of the most conserved elements in alignments of closely related species with a limited number of mismatches (see Note 3).

2.4. Multisequence Conservation of TFBS

The ability to accurately predict active TFBS is a powerful approach for sequence-based discovery of gene regulatory sequences and for elucidating gene regulatory mechanisms (see Note 4). To combat the overabundance of false-positive computational predictions stemming predominantly from the small size of TFBS footprints and from poorly defined position weight matrices (PWM), evolutionary sequence analysis has been proposed as a robust strategy for filtering out false-positive sites (15–18). Methodologically, multiTF is similar to the rVista 2.0 tool (16,17), but implements a different strategy of detecting TFBS present in a multiple alignment. rVista 2.0 works only with pairwise sequence alignments, and requires each site to be present in a short island of high sequence conservation. In contrast, multiTF does not rely on preferential local conservation of functional binding sites vs neutrally evolving background as rVista does, instead it requires a binding site to be present in all the species at the same position as dictated by the alignment. Putative TFBS are identified in all the input sequences by using TRANSFAC PWM matrices to define consensus sequences and the tfSearch utility is used to map these consensus sequences to the genomic sequence of each input species (17,19). MultiTF excludes all TFBS predictions that overlap with exons. Gene annotation for only one of the sequences (the reference sequence) is sufficient to carry out this step. In the final step, multiTF detects TFBS predictions that are shared by all the species and are located at the same position as defined by the alignment. This is achieved by scanning through all the “anchors” or fully conserved nucleotide blocks (nucleotides that are identical in all species in the multiple-sequence alignment; Fig. 9B). If a TFBS from the reference sequence is found to overlap with an “anchor” nucleotide, we project this TFBS position to all the other species by using the alignment and excluding gaps (Fig. 9B). Starting and ending positions of the footprint of the reference sequence TFBS are compared to the starting and ending position for the same TFBS on the same strand as detected by the initial TFBS annotation. If corresponding TFBS can be identified in all the species in the alignment, this is reported by the multiTF.

Open in a separate window

Fig. 9

Transcription factor binding sites can be juxtaposed on the Mulan multiple alignment graphics and several clustering and visualization options are provided for customized analysis (A). Similarly, the binding sites can be visualized within the textual alignment (B).

To analyze Mulan alignments for the presence/absence of conserved TFBS shared among all provided, sequences the user needs to follow these steps:

1. Click on the multiTF button on the summary page (Fig. 4) to forward the alignments to the multiTF program (Fig. 8A).

   

   Open in a separate window

   Fig. 8

   MultiTF portal available from Mulan “summary page.” First menu allows users to define the types of transcription factor binding site matrices to be used in the analysis, along with similarity thresholds (A). Next, the user selects either all or a subset of available transcription factors from the TRANSFAC library (B). Results are provided on an interactive summary page.

2. Upon forwarding to the multiTF analysis initiation page, the user selects from methods and parameters to identify TFBS in individual sequences. First, the user has to choose between the use of the TRANSFAC database of TFBS (http://www.biobase.de/) or user-defined consensus sequences (Fig. 8A).

3. Assuming the most common use of TRANSFAC PWM matrices in description of TFBS binding specificities to scan for binding sites, the user selects the appropriate library of phylogenies (including vertebrate, plant, fungi, nematodes, insects, and bacteria).

4. Two different options are available for detecting TFBS through the use of TRANSFAC libraries. The default option is to use the “optimized for function” search option, which weights individual PWM matrices differently by minimizing and balancing out the abundance of false-negative hits from different matrices. The alternative option is to manually specify matrix similarity cut-off for the annotation of candidate TFBS. The “optimized for function” option utilizes different cut-off parameters for different TFBS, such that no more than three TFBS per 10 kb are predicted in a random sequence (20). Manually selected cut-offs measure sequence similarity to TRANSFAC PWM; the higher the cut-offs are, the fewer sites are predicted.

5. The final option permits the selection of only “high-specificity” matrices in the TFBS annotation. This option subselects a list of TFBS matrices that have <=0.85 cut-off similarity to the TRANSFAC PWM. These are the matrices with the most reliable definitions in the TRANSFAC database.

6. Upon submitting a request, the user is directed to a page that lists all the available transcription factor families alphabetically, where one has to choose the matrices to be used for the analysis by clicking on the provided boxes (Fig. 8B) Alternatively, the user can “select all” to obtain a full repertoire of conserved TFBS.

7. A summary page will comprehensively display the results of the TFBS analysis (Fig. 8C). Here, users can access position and matrix information provided for each sequence independently, as well as the sites can be visualized “on top of” the alignment and used in subsequent clustering analysis (Figs. 8C and ​and9A).9A). The clustering options are similar to the ones available for the rVISTA 2.0 tool, TFBS can be clustered “individually” or “combinatorially” and the sites can be visualized as a summary of conservation (show binding sites by multispecies) or in each sequence individually (show all) (Fig. 9A).