This chapter gives a brief overview of how to use BayesFold.
Before starting BayesFold, align your desired sequences using a program such as ClustalW or Pileup. Generally, alignments will indicate that sequences fall into several families derived from distinct ancestral sequences in the starting pool. Since BayesFold assumes that all the sequences in the input have identical overall structures, select a single family to fold. The underlying Vienna package that generates BayesFold's structural hypotheses deals poorly with gaps, so remove any columns of gaps from the chosen family.
Go to the BayesFold input page at http://bayes.colorado.edu/Bayes and paste in your family of aligned sequences. If the sequences do not include their primers, place those in the primer text boxes. You may also enter a folding temperature and a name for the alignment. Sequences can be entered in FASTA format or in a single-line format where each line contains a label and a sequence separated by spaces and/or tabs. Click validate to ensure that the sequences are in the right format, and then click Continue. On the next input page, correct any mistakes in the labels or sequences, and enter any chemical mappings on the final page. (see Chapter 4, Entering Data for details). Once the alignment has been submitted for folding, BayesFold displays a wait page (refreshed every thirty seconds) indicating when the folding process began and when it was last checked.
As soon as the folding is completed, a results screen will appear. By default, this displays a drawing of the IUPAC consensus sequence folded into the overall best structure. You can easily zoom, pan, or rotate the structure drawing, and use the formatting options to control numbering and end labelling (see Chapter 6, Working with the Structure Drawing). Display any sequence threaded through structure by selecting a different sequence or structure from the pull down menus; the structures are ranked by overall probability, with the most probable structure first in the menu.
If the aligned sequences truly fold into a single structure, then the returned structures should be generally similar. One way to assess this similarity is to enter known motifs of the sequences into the Motifs text box and, using the formatting palette, color the structure drawing to indicate these motifs (see Section 4, “Working with Motifs” and Section 4, “Displaying Data on the Structure Drawing”).Examining the configuration of the motifs in the several likeliest structures will indicate whether they fold similarly or not.
Another way to check that structures are consistent is to use the mismatch data contained in BayesFold's extensive data tables (displayed by clicking the "Show" buttons next to their names). With the best structure selected, open the sequence table and use the display palette to show all mismatch data; then sort that table by the mismatch score. Select the sequence with the largest number of mismatches on the best structure and check its mismatch best index to determine which structure has the fewest mismatches for this outlier sequence. Select that structure and again sort the sequence table by mismatches (see Chapter 7, Working with the Data Tables). If there is a group of sequences that match well with that structure but badly with the best structure, then your sequences probably do not all share the same structure. In this case, you should fold the groups separately by going back to the input screen.
You can also use the tables and the formatting to examine which structures are supported by which kinds of data (including various thermodynamic, chemical mapping, and covariation measures). This enables you to discount any data that is considered suspect. For instance, display mutual information on the drawing and then view several different structures to quickly assess how well this data supports each of them. For a more quantitative determination, open the structure table and examine the changes in each structure's current probability that are caused by alternately displaying and hiding the mutual information data (see Section 3, “Understanding the Probabilities”).
Once you have determined that the sequences truly fold into a single structure and have chosen a set of trusted data, sort the structure table by current probability to determine which structure is the most probable. Large differences in probability between structures (twofold or greater) suggest that one structure is much more probable than another--or, in other words, that the data are definitive. Small differences (e.g. in the second or third decimal place) mean that the structures are about equally likely and therefore additional data are needed to make a rational choice between them. In general, very similar structures that differ by only one or two base pairs have similar probabilities, since there is little evidence to discriminate between them.
When you are finished working with BayesFold, save any desired drawings or data tables, or print them directly from the browser using the Print and Save controls (see Chapter 8, Printing and Saving). BayesFold's graphics are publication quality, and can be modified using any SVG-capable graphics program.