 |
 |
 |
 |
Allan Wilson Centre for Molecular Ecology and Evolution
|
Program Features and Operations
Index:
Available NEXUS blocks
Taxa
Characters
Splits
Compatibility matrix
Lento-Plot
Display of computed network
Unknown block
Named blocks
Back to top
Available NEXUS blocks
When a NEXUS file is first opened in Spectronet, a single window
appears containing a list of the blocks found in the file.
- Double click on a block (or select and press enter) to open it.
- Select a block and press delete to delete it.
- In Spectronet 1.1, all blocks (except comments and blocks having
a type unknown to Spectronet) have unique names. Click the currently
selected block, press F2, or choose Rename Block from the Edit menu
to rename it. Block naming is described in more detail at the end
of this section.
Taxa
Opening a Taxa block produces a window listing the taxa in the file.
- Rename taxa by selecting one, then clicking again, or by pressing
enter.
Characters
Opening a Characters block produces a window displaying the character
sequences.
- Select a site by clicking on the characters themselves.
- Select a taxon by clicking on the name on the left.
- Arrow keys can be used to move selections.
- Choose 'Make Compatibility Matrix' from the Characters menu to compute
and display a compatibility matrix for the character data.
In the Characters menu, 'Get Splits ...' can be used to deduce splits
from the characters, either
- via direct encoding, or
- via Hadamard conjugation.
When generating splits for nucleotide data directly, the "Resolve nucleotides
to R/Y when necessary" option causes Spectronet to deal with sites containing
more than two different character states by converting bases to their
purine/pyrimidine classes. This means that if a site shows both a transition
and a transversion, the statistically more significant transversion will
be recognised. This is achieved by converting the bases A and G to Y (pyrimidine),
and C, T and U to R (purine). If this option is turned off, Spectronet
will simply discard all sites where there are more than two different
character states. This option is enabled by default. You can also adjust
the weight of purine/pyrimidine splits from the default of 1.0.
Note that the free memory display is just a guide - if you choose a block
size larger than the amount of free memory, Windows will free some more
to make room (up to a limit, of course, since Windows itself and the other
programs you want to use at the same time will still have to fit). Also,
the block size actually used is always a power of 2, so choosing a block
size of, say, 30MB, will mean 16MB blocks (the highest power of 2 below
30). If you specify more memory than you actually have, hard disk space
will be used instead, slowing down the program considerably.
Compatibility matrix
The parsimoniously informative characters of the alignment are mapped
onto the rows and columns of this symmetric matrix where local information
is retained, i.e. the characters are not shuffled by the map.
- See the Characters section above for information about the "Resolve
nucleotides to R/Y when necessary" option.
- If a square is black then the corresponding characters are incompatible
(i.e. they represent conflicting information) whereas they are compatible
(i.e. they represent non-conflicting information) if it is white.
- In the Compatibility Matrix window, you can see the pair of sites
a particular square references by briefly resting the mouse cursor
on it. The characters that label the row and the column that intersect
in that square are displayed in a yellow tooltip.
- Copy the image to the clipboard by pressing Ctrl+C or choosing
Copy from the Edit menu. The image that is copied will contain site
number labels along the right and bottom edges of the matrix. These
labels are not displayed on-screen since they are typically too small
to be legible, however the superior resolution of a printer (the usual
destination) means the small size is not a problem. The image is a
metafile (vector) image which can be scaled without degrading the
image quality.
- Two labeled ticks, one along each axis, track the mouse cursor
and display the corresponding site numbers. These are never copied
to the clipboard.
- Adjust the grid cell size and colour using the commands on the
View menu.
Splits
Opening a splits block produces a list of splits, including weight
(i.e. how many characters induced the same split) and information on conflicts
and closeness (i.e. the minimal number of splits needed to transform a
selected split into any one of the other splits of the collection when
walking through the collection's split space).
Select a split.
- The Step column give information on how close other splits are
with the selected split.
- The background color indicates the splits which are compatible
(green) or incompatible (red) with the selected split.
- Double click on a split to see a list of the splits which conflict
with it.
- Press delete to delete the selected splits.
In the Splits menu,
- 'Make Network ...' can be used to make a median network from the
selected splits or all splits.
- 'Make Lentoplot ...' can be used to make a Lentoplot from the selected
splits or all splits.
- 'Make Reduced Split System...' can be used to make a new set of
splits that, when used to create a network with 'Make Network...',
results in a reduced median network as described by H.-J. Bandelt.
The effect of this reduction is to reduce the number of "ladder-like"
reticulations in cases where one evolutionary path is significantly
better supported than all others. This network is consequently easier
to comprehend and manipulate.
It is important to realise that Bandelt's reduction algorithm uses randomised
tie-breaking to determine which of several equally suitable reticulations
to remove. Ties will occur in all but the simplest data sets. When a randomised
tie-breaking operation is required (and it usually is), the user will
be notified via a message window. Often, the random tie-breaking does
not affect, or has a very small effect, on the final network produced,
however no worst-case bounds have been established.
The time this operation takes to run is roughly proportional to the square
of the number of splits in the input split set. It took 8 minutes to reduce
the 439 splits comprising the mt-10.nxs data set on a 933MHz Pentium III
computer with 256Mb of RAM.
Lento-Plot
Opening a Lentoplot displays the plot. Note that the size the plot
appears on screen may be different from that printed - use Print Preview
in the File menu to check.
- Reposition the horizontal axis by dragging it, or double click
to make the area of the graph equal on each side of it.
- Turn the key on or off using the View menu.
- Reposition the key by dragging it.
In the File menu, 'Print...' prints the current Lentoplot ('Print Preview'
and 'Print Setup...' are also enabled).
In the Edit menu, 'Copy' copies the current Lentoplot to the clipboard
as a Windows Metafile, which can then be pasted into applications such
as Microsoft Word.
Display of computed network
The manipulation of the displayed network is modeled after the manipulation
of splits graphs produced by the SplitsTree
package.
If the option 'Auto Arrange' in the 'Make Reduced Network' dialog box
is ticked, as it is by default, the computed median network is initially
arranged in such a manner as to minimise the possibility of near-parallel
split angles occurring, and in so doing causing confusion about which
edges relate to which splits. The auto-arrange feature will generally
produce geometrically similar networks from networks having similar topology
and split weights, so it is also useful for comparing networks by eye.
(It is possible to waste a lot of time manually arranging two very similar
networks for visual comparison, only to find you have given them totally
different layouts.) The auto-arrange algorithm works by ranking the splits
by their weights, setting the edges of the first ("heaviest") split to
lie horizontally, and rotating the set of edges for each successive split
by an angle theta = pi/(1+sqrt(5)). This value for theta causes the resulting
split angles to be maximally spread when the number of splits is large.
You can (and probably should) still manually rearrange the network after
using 'Auto Arrange' - it is useful as a starting point, but rarely will
the algorithm give you a balanced-looking network with few edge crossings.
If 'Auto Arrange' is unticked, the generated network will initially appear
as one or more vertical lines interspersed with taxa labels. "Unfold"
the network by successively clicking on and dragging all vertical sections
of the folded-up network sideways. A good rule of thumb for expanding
the network completely is, "If you see a vertical edge, drag it sideways."
Once there are no more vertical edges, you can be fairly certain that
there are no more edges hidden beneath other edges.
Further options for manipulating the network are:
- Drag splits and labels around with the mouse.
- Select multiple items by holding down shift and clicking (this
can also be used to de-select items).
- Modify the appearance of the selected items by double clicking,
or choosing 'Properties...' from the Edit menu or the context (right
click) menu.
- Rotate a split and all other splits on one side of it by holding
down control and dragging the split (very useful for tree-like structures).
- Rotate the entire network by holding control and dragging an empty
area of the window.
- Toggle visibility of a scale in either the View menu or the context
menu.
- Drag the scale around, or resize it at either end.
- Rotate the scale by holding down control and dragging it.
- Toggle whether edges are displayed with equal lengths in either
the View menu or the context menu.
- Select entire classes of item from either the Edit menu or the
context menu.
- Add a length label to an edge by right clicking on that edge and
choosing 'Edge Label'.
In the File menu, 'Print...' prints the current graph ('Print Preview'
and 'Print Setup...' are also enabled).
In the Edit menu, 'Copy' copies the current graph to the clipboard as
a Windows Metafile, which can then be pasted into applications such as
Microsoft Word.
Unknown blocks
Opening a block that is unknown to Spectronet shows the text form
of that block.
- Block types that Spectronet does not understand can be viewed and
edited in their text form.
- It is not possible to rename a block of unknown type - although
it appears to take effect when you edit the block name in the main
window, the name will not be saved and will revert to a sequence-number-based
name when the file is reloaded.
Comments between blocks are treated similarly to unknown blocks, and can
be edited.
Named blocks
It is the intention of the authors to add new functionality to Spectronet
in the near future that will exploit relationships between NEXUS blocks:
for example, it would be useful to be able to select a split in the Splits
window and have the Characters window immediately update to show the characters
that induce that split. Or, to select an edge in a median network graph
window and have the Splits window automatically select the corresponding
split. This kind of dynamic updating requires that NEXUS blocks "know"
what other blocks they are connected to, and this requires that the blocks
have unique names.
Spectronet now gives all the blocks in a NEXUS file unique names, which
are saved and reloaded using a comment field that is invisible to other
programs (see the NEXUS format page for more details). The user can easily
rename blocks in the main NEXUS file window (see the section, "Available
NEXUS blocks"). Block windows now have titles of the form <NEXUS filename>:<blockname>,
e.g. "matK.nxs:ST_Splits_42". These names also display in the Windows
menu, so it is easier to navigate when you have many files or blocks open.
If a block in a file does not contain the special naming comment, it is
given a name of the form <block_type>_<n>, where n is just
a serial number that increments up from 1, e.g. "ST_Splits_42". This means
all current NEXUS files will work just fine.
|
