Content uploaded by Andrew S. Harvey
Author content
All content in this area was uploaded by Andrew S. Harvey on Jan 20, 2014
Content may be subject to copyright.
1
ClustalG: Software for analysis of activities and sequential events
Clarke Wilson
Canada Mortgage and Housing Corporation
C5-318 700 Montreal Road
Ottawa, Canada K1A 0P7
Phone 613 748 4670
Fax 613 748 4865
email cwilson@cmhc-schl.gc.ca
Andrew Harvey
Department of Economics
St. Mary’s University
Halifax, Nova Scotia B3H 3C3
Phone 902 420 5676
email andrew.harvey@stmarys.ca
Julie Thompson
Intitute de Genetique et de Biologie Moleculaire et Cellulaire
Strasboug, France
julie@igbmc.u-strasbg.fr
1999
Paper presented at the Workshop on Longitudinal Research in Social Science: A Canadian Focus,
Windermere Manor, London, Ontario, Canada, October 25-27, 1999.
2
ABSTRACT
The paper describes a new software package for sequence alignment analysis called ClustalG. The
package is a rewrite of the well-known Clustal series of alignment packages. The main new
feature of ClustalG is the recognition of input word sequences of up to six characters. This
effectively eliminates the 20 letter constraint implmented in biological software on the number of
event categories available to the researcher.
The essential ClustalG windows are shown and the meaning of the most common variable settings
are discussed. Some elementary alignments and clustering trees are illustrated. The software
package including the help file and a number of time use sample files is freely available to any user
from the St. Mary’s University ftp site.
Key words: activity patterns sequence alignment software
Acknowledgements
The authors acknowledge the original work on the Clustal program group by Des Higgins and
Paul Sharp, and we thank Toby Gibson of the European Molecular Biology Laboratory for
agreeing to allow us to redesign the program for use outside molecular biological applications.
Their work represents a huge investment of research funds and talent that will now be applied to
subjects far beyond those for which Clustal was originally designed.
This work is part of the Activity Settings, Sequencing and the Measurement of Time Allocation
Patterns, project funded by the Social Science and Humanities Research Council of Canada.
1. The ClustalG project
The Activity Settings: Design, Measurement, and Analysis research project funded by the Social
Science and Humanities Research Council of Canada in 1994 produced a number of papers and
publications that have illustrated the application of sequence alignment methods and software as
developed in molecular biology to time use and transportation research [1, 2, 3]. These have all
used versions of the Clustal programs maintained at the European Molecular Biology Laboratory.
The results of the applications suggest that alignment methods hold great promise for examining
social processes that consist of sequences of activities. Abbott [4] has reviewed a variety of
research into sequential processes based on alignment or optimal matching as the methods are
sometimes called. However, the biological software contains a number of features that have no
place in social science research, and available packages generally limit the eligible alphabet to just
over 20 characters.
A subsequent SSHRCC project, Activity Settings, Sequencing and the Measurement of Time
Allocation Patterns, has contracted the Clustal programmer, Julie Thompson, to amend the
windows version, ClustalX, for the research in any discipline that deals with sequential processes.
3
The product is called ClustalG (for general) and is available from the ftp site at St. Mary’s
University, Halifax.
The properties of ClustalW and ClustalX have been published [5,6]. Briefly, the packages
implement a two stage process of calculating the pairwise similarities in a set of sequences then
constructing a tree from transformations of the similarities. The tree is used to guide the
progressive multiple alignment of the set of sequences.
ClustalG has deleted the explicitly biochemical features of ClustalX, has expanded the input
routines to accept multiple letter words of up to six characters, and has created an new output file
that specifies the members of each step by which the program clusters individuals into
progressively larger and more general pattern groupings. The key feature is the introduction of
multiple letter words because this permits analysts to use complex coding schemes that are usual
in many sciences. Analysts may use different positions in the word to indicate different dimensions
of events. In our example data, the first two positions indicate an activity, the third indicates
location, and the fourth who else was present.
2. Sequence alignment methods
Sequence alignment, or optimal string matching as the methods are also called, employ
combinatorial algorithms to calculate measures of either similarity or distance between character
sequences. See Waterman [7] for a comprehensive treatment of alignment mathematics and
biological applications. When stages of processes or activities are represented by characters, these
measurements can form the basis of taxonomies of the behaviour being examined. Alignment
methods provide the most rigorous basis available for classifying groups of character sequences.
The general process can be illustrated by writing the elements of two sequences in the margins of
a comparison table and placing an asterisk in cells for which marginal elements match. Consider
the comparison of letters of [mississippi] and [missouri] shown in Figure 1.
Figure 1: Comparison table for [mississippi] and [missouri]
m i s s i s s i p p i
m *
i * * * *
s * * * *
s * * * *
o
u
r
i * * * *
The degree similarity of the two names is established in the first syllable as shown by the
downward sloping diagonal pattern of stars. The [iss] substring is repeated in [mississippi] and
4
this is illustrated by the second diagonal, translated three positions to the right. The remaining
letter matches are more or less random.
The alignment algorithms are based on calculation of a cumulative score beginning at the upper
left cell and proceeding to the lower right. A cell’s score is based on the preceding score plus its
own value. Values are determined by weighting systems related to the substance of the problem in
question. A path can be found that leads backwards from the lower right cell through the highest
value cells to the upper left. The order in which letters are included in the path, and in particular
whether a letter matches another letter or is placed against a gap, determines the pairwise
alignment. Gaps may be inserted in either sequence to allow identical letters to match. Optimal
paths and alignments are often not unique. One option for the alignment of [mississippi] and
[missouri] is shown below:
m i s s - - - i s s i p p i
m i s s o u r i - - - - - -
The exact patter of letters in positions five and following is determined by the system of scoring
weights and gap penalties used.
Pairwise alignment may be generalized to multiple alignments by defining comparison tables and
paths in N dimensions. However, for N greater than about 10 sequences, the algorithms are
prohibitively costly in time and memory space. Multivariate alignments are usually implemented
using approximate methods based on pairwise measures. This is the case with the Clustal program
family.
3. ClustalG screen
The ClustalG screen is shown in Figure 1 as it is displayed when the program is executed. Seven
menus control the loading of sequence files, editing, alignments, preparation of trees calculated as
a result of the clustering process performed progressively on the sequence file, coloring, depiction
of special sequences or segments (quality), and the help screens. This presentation deals only with
file manipulation and specification of alignment parameters. The ClustalG online help facility
covers the other items.
ClustalG operates in Multiple Alignment or Profile Alignment modes, which are selected from the
first of the drop down boxes. Multiple Alignment mode uses a single screen. Profile Alignment
mode uses two screens because a profile alignment is an alignment of two previously constructed
alignments. Multiple alignment mode is normally used first to find useful arrangements of
sequence data. The researcher may later want to combine various alignments.
The Alphabet Size from one to six characters must be chosen before sequences are loaded in to
ClustalG. All elements of all sequences are treated as having a constant size.
The Windows menu bar at the bottom is not part of the ClustalG screen.
5
4. File menu options
6
Load sequences:
This is the first step and is mandatory. Selection of the load sequence option invokes a Windows
Open screen that allows user to specify drive, folder and filename. Sequence labels are written in
the left-hand box and the sequence elements are written to the right.
Many single letter sequence formats are used in biology. ClustalG allows all that have been
implemented in ClustalX in addition to the multiple letter words. The simplest is the Pearson or
Fasta format which has been used in the example files. This format begins each sequence record
with a greater-than symbol and the characters on the line following are treated as a label. Lines
following the first line are treated as sequence data and are read until another greater-than symbol
or the end of file character is found. For example:
> 1346wda 12e 12
rewaeawreamr
> 1444sna 17e 17
rrkcdecdacemckkmr
An example of a sequence that uses 4 letter words is:
> 2011mna 15e 15
ZzhaPchaPchaTrtaWkwaWkwaWkwaTrtaZzhaTvhaEthfZzhaZzhaFchfZzha
The diary reads: asleep, home, alone; personal care, home, alone; personal care, home, alone;
travel, location is travel, alone; work, at workplace, alone; etc...
Append sequences:
Additional sequences can be added to a file previously loaded.
Save sequences as:
Permits user to specify a new file location for edited sequence files or for new alignments using
different sets of parameter values.
Profile options:
Similar to the multiple alignment options except that two files are specified
Write as Postscript:
Creates Postscript graphic output file.
7
5. Alignment menu options
The first subgroup of options launches and controls the output of ClustalG files, including the
alignment file (*.aln), the dendrogram file (*.dnd). The second group allows amendment of
existing alignments, and the third group sets the parameters for the alignment. Group three is
described first.
Reset These are selected with the mouse and display a check when selected
Save log file This is also selected with the mouse and controls the writing of the
pairwise similarity values (*.lg1) and the grouping steps (*.lg2) to disk.
Pairwise alignment parameter screen
This option invokes a dialogue box that allows the user to control the set of pairwise alignments
that are computed first and from which the guide tree is calculated which in turn controls the
multiple alignment step. Gap open and gap extension parameters should be set.
A file containing weights for element matches and substitutions may be input. The default is to use
an identity matrix with diagonal values set to 10. Gap penalty parameters should relate to weight
matrix values.
8
Multiple Alignment Parameter screen
This options invokes another dialogue box the allows the user to control the set of parameters
used by ClustalG in conjunction with pairwise similarity scores to calculate a guide tree, and from
there to assemble the multiple alignment. The weight matrix usage is the same as for the pairwise
screen.
Gap parameter option screen
A dialogue box allows further specification of gap parameters.
Output format option screen
User selects one or more formats for the output files
The line width may be set to control alignment appearance. The alignment is written as a series of
blocks of fixed width. Where output lines are comparatively short, they may fit on letter or legal
paper in portrait or landscape orientation. Where lines are too long the user can control block
width up to 1000 characters.
9
The aligned sequences may be written in input order or in an order that roughly follows their
grouping order.
The user may choose to have the
parameter set written to a file for
future reference.
Do Complete Alignment
Screen
This screen allows the
user to name the output
alignment and
dendrogram files. The
default is to use the same
name as the sequence file
that was loaded with extensions of *.aln and *.dnd. An alignment file is shown later. The guide
tree or dendrogram is written as a text file of nested parentheses containing sequence labels and
the branch lengths of the tree which can be drawn from the nesting pattern. No tree is drawn.
However, the file format is recognized by biological graphics software.
10
Produce guide tree only
This generates only the *.dnd file. This may be used with tree drawing software (for example
Treeview by Rod Page, University of Glasgow, or Phylip by Joe Felsenstein , University of
Washington) to display the dendrogram graphically. A Treeview [8] screen is shown later.
Alignment from guide tree screen
Specifies an existing guide tree to control the multiple alignment. New multiple parameters may
be selected.
6. ClustalG Alignment Screen
ClustalG has identified three primary behavioural groups in the test data file. Their multiple
alignment is shown above and the groupings are shown in the tree diagram on the next page. The
order of output in the alignment is roughly but not precisely that in the tree diagram and the two
should be used together to identify the membership of the primary behavioural groupings that
occur in the alignment.
11
7. Treeview illustration of the ClustalG guide tree file
The tree identifies similar groups of sequences precisely. The alignment describes what the
activity patterns are. The middle group in the tree diagram are employed people who had several
work episodes in their diaries.
12
References
1. Wilson W.C. 1998 Activity pattern analysis by means of sequence alignment methods,
Environment and Planning, volume 30, pp. 1017-1038
2. Wilson W.C. 1998 Analysis of travel behaviour using sequence alignment methods,
Transportation Research Record, number 1645, pp. 52-59.
3. Harvey A.S. and Wilson W.C. 1998, Evolution of daily activity patterns: a study of the Halifax
panel survey, paper presented at Thematic Group 1, Time-Use, World Congress of Sociology (in
conjunction with Association, International Association for Time Use Research) University of
Quebec, Montreal, July 26-August 1, 1998.
4. Abbott A. 1999 the review paper. citation to come
5. Thompson,J.D., Gibson,T.J., Plewniak,F., Jeanmougin,F. and Higgins,D.G.1997, The
ClustalX windows interface: flexible strategies for multiple sequence alignment aided by quality
analysis tools. Nucleic Acids Research, 24:4876-4882.
6. Thompson, J.D., Higgins, D.G. and Gibson, T.J. 1994 CLUSTAL W: improving the sensitivity
of progressive multiple sequence alignment through sequence weighting, positions-specific gap
penalties and weight matrix choice. Nucleic Acids Research, 22:4673-4680.
7. Waterman, M. 1995, Introduction to Computational Biology, Chapman and Hall, London
8. Page, R. D. M. TREEVIEW: An application to display phylogenetic trees on personal
computers, Computer Applications in the Biosciences, 12:357-358.
