Spectrum 2.0 User's Guide

Software by Michael Charleston and Rod Page
Guide by Rod Page

This user's guide describes the operation of the new version of Spectrum, and assumes that you are familiar with the basics of spectral analysis.

Contents

  • Getting started
  • Getting spectra
  • Looking at spectra
  • Getting a tree
  • Output of results
  • Saving your data file
    •

    Getting started

    Installation

    Macintosh

    For the Macintosh versions (68K and PPC) there is very little installation necessary: simply extract the program applications and place them where you want on your hard disc. Since they come as binhex'd self-extracting archives, this will be automatic on most machines.

    In order to correctly display trees in the log window, Spectrum uses a font called "COMPONENTMonaco." To use this font, drop the "COMPONENT Font File" onto your System folder.

    IBM PC

    The IMB PC version (32 bit only) is stored as the file SPECTRUM.ZIP. Extract this file into a temporary folder or directory . Among the resulting files is SETUP.EXE. This program will install SPECTRUM for Win32 on your computer. If you are running Windows 95 the installation program will add SPECTRUM to your registry so that files with the extension .SPC will have the SPECTRUM document icon when viewed with Expolorer and in the Open and Save dialog boxes in SPECTRUM. The installation program also supports the Uninstall feature of Windows 95, so that you can remove SPECTRUM from your computer using the Add/Remove Programs control panel application.

    Using Spectrum

    Spectrum is a standard Macintosh program with menus and dialog boxes. Once a data file has been opened the program displays a log window whose title bar shows the name of the data file. Information about the status of the program, and results of analyses are displayed in this log window. Other windows can be opened to graphically display spectra and trees.

    Reading data files

    Spectrum uses standard NEXUS format used by programs such as PAUP, MacClade and SplitsTree. The program should read most standard NEXUS files, but does not support all the NEXUS commands or blocks (the program will display a warning if it has ignored a command or block). If you have difficulties, and the file can be read by either PAUP or MacClade, then please contact the program's authors.

    Importing PHYLIP files

    Spectrum has limited support for PHYLIP format files. The Import PHYLIP file command on the File menu displays a dialog box in which you specify the type of PHYLIP file you wish to import:

    Spectrum can import both interleaved and sequential format PHYLIP files. However it cannot read files that have options such as "W" (weights).
    Spectrum will read distance matrices in either lower-left, upper-right, or square formats. However, it will not handle replicates, or other options.
    Once it has successfully read the PHYLIP file, Spectrum will open an untitled log window.

    Getting spectra

    If you have discrete binary character data then you can calculate a spectrum directly from that data. If you have nucleotide sequences then you must either specify how you want the four nucleotides mapped onto two states, or you can compute a pair-wise distance matrix and compute a spectrum from those distances.

    Mapping nucleotides

    In order to convert four-state characters to two-state characters you must specify a mapping. By default, Spectrum converts nucleotide data to purines and pyrimidines. You can change this using the Mapping command. This command displays the following dialog box:

    The Average over all option extracts splits by averaging over all the possible mappings, and allows you to specify the weights for each possible mapping. The default is to treat each mapping equally.

    Getting spectra from distances

    If you have DNA or RNA sequence data, Spectrum can compute a range of measures of sequence distance. These are:
    Hamming The proportion of sites at which the two sequences differ.
    Jukes-Cantor Assumes equal base frequencies and no transition/transversion bias.
    Kimura 2-parameter Assumes equal base frequencies and allows a transition/transversion bias.
    Tamura-Nei Allows for unequal base frequencies and a transition/transversion bias. Similar to the HKY85 model, but Tamura-Nei allows the two types of transition to occur at different frequencies.
    LogDet Designed to overcome the effects of unequal base composition.
    Nucleotide composition Not intended a measure of evolutionary distance, but simply how similar the sequences in their base composition.
    When computing the pair-wise distance between two sequences, Spectrum ignores any site for which one or other or both sequences have a missing value.

    Once the distance matrix is calculated, a distance spectrum can be constructed. If the distances are additive, then the distance spectrum will contain the edge lengths of the underlying tree; if the distances deviate from additivity then there will be some other splits with positive support values.

    Looking at spectra

    You have two ways of looking at the spectrum:

    Lento plots

    A Lento plot shows the support (and optionally the conflict) for each split in the spectrum, above the current threshold. The Plot Options command specifies how your plot will look. The options are:
    Rank the splits by their support, and display each split's conflict below that split. This is the default option
    Rank the splits by their support; conflicts are not shown.
    The splits are plotted in the order of their split number, and display the conflict for each split.
    The splits are plotted in the order of their split number; conflicts are not shown.

    If you are going to display the split conflict values, you can choose to normalise the conflicts by the number of conflicting splits, or by the number of splits that could possibly conflict with that split.

    If you chose the Plot command the program will display the spectrum in a separate chart window:

    The splits are coloured light grey if they correspond to terminal taxa, and otherwise are white. If you have computed a tree then the splits in that tree will be coloured black. Pendant splits are labelled by the corresponding taxon name; internal splits are labelled by their split number.

    Hints

    Lists

    The List command will list the splits above the current threshold. It will respect the options for Lento plots.

    Looking at individual splits

    You can get the support and conflict of individual splits by using the Evaluate bipartition and Evaluate split commands. These do exactly the same thing, however the Evaluate split command requires you to know the number of the split you want to look at. If you don't know the number, you can use the Evaluate bipartition command, which displays the following dialog box:

    By moving taxa from the left-hand list to the right-hand list you create a split. The corresponding split number is displayed at the top of the dialog box. In this example the split {Cassowary, Emu} is shown, which is split 12.

    Getting a tree

    Spectrum computes both closest (not yet) and Manhattan trees. Once you have a tree you can display the tree in the log window using the Show tree and commands, or you can view the tree in a separate tree window:

    The toolbar at the top of the tree window has the following buttons to control how the tree is displayed.

    Display the tree as an unrooted tree. This is the default option
    Display the tree as a rooted tree (cladogram).
    Display the tree as a rooted tree with branch lengths proportional to split support.
    Toggle on and off the display of the split number each edge in the tree corresponds to.

    Saving the tree

    You can save the tree displayed in the tree window to a NEXUS format tree file using the Save tree command.

    Assessing how well supported the tree is

    You can evaluate support for the edges in the current tree using the Plot NNI Neighbourhood command.

    This will display a chart of the support for each internal edge in the tree, and the two edges that correspond to the two nearest neighbour interchanges (NNI's) around that edge. In other words, you can see whether making a given edge is strongly supported relative to a local rearrangement of that tree.

    In the NNI plot, each internal edge in the tree is plotted, followed by the two NNI edges. For example, in the plot below

    split 12 is in the tree, and splits 10 and 6 are the two splits we obtain by interchanging the nearest neighbours around split 12. In this example, there is rather less support for 10, and very little for 6. If either of the alternative splits had similar support to the split in the tree then that edge could be judged poorly supported.

    Output of results

    The output from any window can be printed or copied to the Clipboard. In addition, the Lento plots and trees can be saved as graphics files (PICT format) for editing by a graphics program.

    You can also save your output for use by other programs. For example, a spectrum can be saved as a tab-delimited text file for importing into a spreadsheet program such as Excel. Any trees you have found can be saved as NEXUS tree files which can be viewed by TreeView, and read into programs that support NEXUS files such as PAUP and MacClade.

    Saving your data file

    Spectrum can save your data to a NEXUS format file that has the following icon:

    This file contains the following:

    Warning

    Spectrum only saves TAXA, CHARACTERS, DISTANCE, and TREES blocks. Any other blocks in your file will not be saved. Hence, you may wish to save the file under another name so as not to lose any information in your original data file.