CLUSTAL-OMEGA is a general purpose multiple sequence alignment program
for proteins.



INTRODUCTION

Clustal-Omega is a general purpose multiple sequence alignment (MSA)
program for proteins. It produces high quality MSAs and is capable of
handling data-sets of hundreds of thousands of sequences in reasonable
time.

In default mode, users give a file of sequences to be aligned and
these are clustered to produce a guide tree and this is used to guide
a "progressive alignment" of the sequences.  There are also facilities
for aligning existing alignments to each other, aligning a sequence to
an alignment and for using a hidden Markov model (HMM) to help guide
an alignment of new sequences that are homologous to the sequences
used to make the HMM.  This latter procedure is referred to as
"external profile alignment" or EPA.

Clustal-Omega uses HMMs for the alignment engine, based on the HHalign
package from Johannes Soeding [1].  Guide trees are optionally made
using mBed [2] which can cluster very large numbers of sequences in
O(N*log(N)) time.  Multiple alignment then proceeds by aligning larger
and larger alignments using HHalign, following the clustering given by
the guide tree.

In its current form Clustal-Omega can only align protein sequences but
not DNA/RNA sequences. It is envisioned that DNA/RNA will become
available in a future version.



SEQUENCE INPUT:

-i, --in, --infile={<file>,-} 
	Multiple sequence input file (- for stdin)

--hmm-in=<file>           
	HMM input files

--dealign                 
	Dealign input sequences

--profile1, --p1=<file>   
	Pre-aligned multiple sequence file (aligned columns will be kept fixed)

--profile2, --p2=<file>   
	Pre-aligned multiple sequence file (aligned columns will be kept fixed)


For sequence and profile input Clustal-Omega uses the Squid library
from Sean Eddy [3].


Clustal-Omega accepts 3 types of sequence input: (i) a sequence file
with un-aligned or aligned sequences, (ii) profiles (a multiple
alignment in a file) of aligned sequences, (iii) a HMM. Valid
combinations of the above are:

(a) one file with un-aligned or aligned sequences (i); the sequences
    will be aligned, and the alignment will be written out. For this
    mode use the -i flag. If the sequences are aligned (all sequences
    have the same length and at least one sequence has at least one
    gap), then the alignment is turned into a HMM, the sequences are
    de-aligned and the now un-aligned sequences are aligned using the
    HMM as an External Profile for External Profile Alignment (EPA).
    If no EPA is desired use the --dealign flag.

    Use the above option to make a multiple alignment from a set of
    sequences. A sequence file must contain more than one sequence (at
    least two sequences).

(b) two profiles (ii)+(ii); the columns in each profile will be kept
    fixed and the alignment of the two profiles will be written
    out. Use the --p1 and --p2 flags for this mode.

    Use this option to align two alignments (profiles) together.

(c) one file with un/aligned sequences (i) and one profile (ii); the
    profile is converted into a HMM and the un-aligned sequences will
    be multiply aligned (using the HMM background information) to form
    a profile; this constructed profile is aligned with the input
    profile; the columns in each profile (the original one and the one
    created from the un-aligned sequences) will be kept fixed and the
    alignment of the two profiles will be written out. Use the -i flag
    in conjunction with the --p1 flag for this mode. 
      The un/aligned sequences file (i) must contain at least two
    sequences. If a single sequence has to be aligned with a profile
    the profile-profile option (b) has to be used.

    Use the above option to add new sequences to an existing
    alignment.

(d) one file with un-aligned sequences (i) and one HMM (iii); the
    un-aligned sequences will be aligned to form a profile, using the
    HMM as an External Profile. So far only one HMM can be input and
    only HMMer2 and HMMer3 formats are allowed. The alignment will be
    written out; the HMM information is discarded. As, at the moment,
    only one HMM can be used, no HMM is produced if the sequences are
    already aligned. Use the -i flag in conjunction with the --hmm-in
    flag for this mode. Multiple HMMs can be inputted, however, in the
    current version all but the first HMM will be ignored.

    Use this option to make a new multiple alignment of sequences from
    the input file and use the HMM as a guide (EPA).


Invalid combinations of the above are: 

(v) an un/aligned sequence file containing just one sequence (i)

(w) an un/aligned sequence file containing just one sequence and a profile
    (i)+(ii)

(x) an un/aligned sequence file containing just one sequence and a HMM
    (i)+(iii)

(y) two or more HMMs (iii)+(iii)+... cannot be aligned to one another.

(z) one profile (ii) cannot be aligned with a HMM (iii)


The following MSA file formats are allowed: 

    a2m=fasta, (vienna)
    clustal,
    msf,
    phylip,
    selex,
    stockholm


Prior to MSA, Clustal-Omega de-aligns all sequence input (i). However,
alignment information is automatically converted into a HMM and used
during MSA, unless the --dealign flag is specifically set.  Profiles
(ii) are not de-aligned.

The Clustal-Omega alignment engine can at the moment not process
DNA/RNA. If a sequence input file (i) or a profile (ii) is interpreted
as DNA/RNA the program will terminate during the file input stage.



CLUSTERING:

  --distmat-in=<file>       
	Pairwise distance matrix input file (skips distance computation)

  --distmat-out=<file>      
	Pairwise distance matrix output file

  --guidetree-in=<file>     
	Guide tree input file 
	(skips distance computation and guide tree clustering step)

  --guidetree-out=<file>    
	Guide tree output file

  --mbed                    
	Fast, Mbed-like clustering for guide tree calculation

  --mbed-iter               
	Use Mbed-like clustering also during iteration
                            

In order to produce a multiple alignment Clustal-Omega requires a
guide tree which defines the order in which sequences/profiles are
aligned. A guide tree in turn is constructed, based on a distance
matrix. Conventionally, this distance matrix is comprised of all the
pair-wise distances of the sequences. The distance measure
Clustal-Omega uses for pair-wise distances of un-aligned sequences is
the k-tuple measure [4], which was also implemented in clustal1.83 and
clustalW2 [5,6].  If the sequences inputted via -i are aligned
Clustal-Omega uses the Kimura-corrected pairwise aligned identities
[7]. The computational effort (time/memory) to calculate and store a
full distance matrix grows quadratically with the number of
sequences. Clustal-Omega can improve this scalability to N*log(N) by
employing a fast clustering algorithm called mBed [2]; this option is
invoked by specifying the --mbed flag.

Clustal-Omega uses Muscle's [8] fast UPGMA implementation to construct
its guide trees from the distance matrix. By default, the distance
matrix is used internally to construct the guide tree and is then
discarded. By specifying --distmat-out the internal distance matrix
can be written to file.  The guide trees by default are used
internally to guide the multiple alignment and are then discarded. By
specifying the --guidetree-out option these internal guide trees can
be written out to file. Conversely, the distance calculation and/or
guide tree building stage can be skipped, by reading in a
pre-calculated distance matrix and/or pre-calculated guide tree. These
options are invoked by specifying the --distmat-in and/or
--guidetree-in flags, respectively. However, distance matrix reading
is disabled in the current version. By default, distance matrix and
guide tree files are not over-written, if a file with the specified
name already exists. In this case Clustal-Omega aborts during the
command-line processing stage. To force over-writing of already
existing files use the --force flag (see MISCELLANEOUS).

Guide trees can be iterated to refine the alignment (see section
ITERATION). Clustal-Omega takes the alignment, that was produced
initially and constructs a new distance matrix from this
alignment. The distance measure used at this stage is the Kimura
distance [7]. By default, Clustal-Omega constructs a full distance
matrix at this stage, which will then be used to create an improved
(iterated) new guide tree. To use mBed like clustering at this stage
the --mbed-iter flag has to be set. While Kimura distances in general
are much faster to calculate than k-tuple distances, time and memory
requirements still scale quadratically with the number of sequences
and mBed-type clustering should be considered for large cases.



ALIGNMENT OUTPUT:

  -o, --out, --outfile={file,-} Multiple sequence alignment output file (default: stdout)

  --outfmt={a2m=fa[sta],clu[stal],msf,phy[lip],selex,st[ockholm],vie[nna]} MSA output file format (default: fasta)
                            

By default Clustal-Omega writes its results (alignments) to stdout. An
output file can be specified with the -o flag. Output to stdout is not
possible in verbose mode (-v, see MISCELLANEOUS) as verbose/debugging
messages would interfere with the alignment output.  By default,
alignment files are not over-written, if a file with the specified
name already exists. In this case Clustal-Omega aborts during the
command-line processing stage. To force over-writing of already
existing files use the --force flag (see MISCELLANEOUS).

Clustal-Omega can output alignments in various formats by setting the
--outfmt flag:

  * for Fasta format set: --outfmt=a2m  or  --outfmt=fa  or  --outfmt=fasta

  * for Clustal format set: --outfmt=clu  or  --outfmt=clustal

  * for Msf format: set --outfmt= msf

  * for Phylip format set: --outfmt=phy  or  --outfmt=phylip

  * for Selex format set: --outfmt=selex

  * for Stockholm format set: --outfmt=st  or  --outfmt=stockholm

  * for Vienna format set: --outfmt=vie  or  --outfmt=vienna


ITERATION:

  --iterations, --iter=<n>  Number of (combined guide tree/HMM) iterations

  --max-guidetree-iterations=<n> Maximum guide tree iterations

  --max-hmm-iterations=<n>  Maximum number of HMM iterations
                            

By default, Clustal-Omega calculates (or reads in) a guide tree and
performs a multiple alignment in the order specified by this guide
tree. This alignment is then outputted. Clustal-Omega can 'iterate'
its guide tree. The hope is that the (Kimura) distances, that can be
derived from the initial alignment, will give rise to a better guide
tree, and by extension, to a better alignment.

A similar rationale applies to HMM-iteration. MSAs in general are very
'vulnerable' at their early stages. Sequences that are aligned at an
early stage remain fixed for the rest of the MSA. Another way of
putting this is: 'once a gap, always a gap'. This behaviour can be
mitigated by HMM iteration. An initial alignment is created and turned
into a HMM. This HMM can help in a new round of MSA to 'anticipate'
where residues should align. This is using the HMM as an External
Profile and carrying out iterative EPA.  In practice, individual
sequences and profiles are aligned to the External HMM, derived after
the initial alignment. Pseudo-count information is then transferred to
the (internal) HMM, corresponding to the individual
sequence/profile. The now somewhat 'softened' sequences/profiles are
then in turn aligned in the order specified by the guide
tree. Pseudo-count transfer is reduced with the size of the
profile. Individual sequences attain the greatest pseudo-count
transfer, larger profiles less so. Pseudo-count transfer to profiles
larger than, say, 10 is negligible. The effect of HMM iteration is
more pronounced in larger test sets (that is, with more sequences).

Both, HMM- and guide tree-iteration come at a cost of increasing the
run-time. One round of guide tree iteration adds on (roughly) the time
it took to construct the initial alignment. If, for example, the
initial alignment took 1min, then it will take (roughly) 2min to
iterate the guide tree once, 3min to iterate the guide tree twice, and
so on. HMM-iteration is more costly, as each round of iteration adds
three times the time required for the alignment stage. For example, if
the initial alignment took 1min, then each additional round of HMM
iteration will add on 3min; so 4 iterations will take 13min
(=1min+4*3min). The factor of 3 stems from the fact that at every
stage both intermediate profiles have to be aligned with the
background HMM, and finally the (softened) HMMs have to be aligned as
well. All times are quoted for single processors.

By default, guide tree iteration and HMM-iteration are coupled. This
means, at each iteration step both, guide tree and HMM, are
re-calculated. This is invoked by setting the --iter flag. For
example, if --iter=1, then first an initial alignment is produced
(without external HMM background information and using k-tuple
distances to calculate the guide tree). This initial alignment is then
used to re-calculate a new guide tree (using Kimura distances) and to
create a HMM. The new guide tree and the HMM are then used to produce
a new MSA.

Iteration of guide tree and HMM can be de-coupled. This means that the
number of guide tree iterations and HMM iterations can be
different. This can be done by combining the --iter flag with the
--max-guidetree-iterations and/or the --max-hmm-iterations flag.  The
number of guide tree iterations is the minimum of --iter and
--max-guidetree-iterations, while the number of HMM iterations is the
minimum of --iter and --max-hmm-iterations.  If, for example, HMM
iteration should be performed 5 times but guide tree iteration should
be performed only 3 times, then one should set --iter=5 and
--max-guidetree-iterations=3. All three flags can be specified at the
same time (however, this makes no sense). It is not sufficient just to
specify --max-guidetree-iterations and --max-hmm-iterations but not
--iter. If any iteration is desired --iter has to be set.


LIMITS (will exit early, if exceeded):

  --maxnumseq=<n>           Maximum allowed number of sequences

  --maxseqlen=<l>           Maximum allowed sequence length

Limits can be imposed on the number of sequences in the input file
and/or the lengths of the sequences. This cap can be set with the
--maxnumseq and --maxseqlen flags, respectively. Clustal-Omega will
exit early, if these limits are exceeded.
                            

MISCELLANEOUS:

  --auto                    Set options automatically (might overwrite some of your options)

  --threads=<n>             Number of processors to use

  -h, --help                Print help and exit

  -v, --verbose             Verbose output (increases if given multiple times)

  --version                 Print version information and exit

  --long-version            Print long version information and exit

  --force                   Force file overwriting


Users may feel unsure which options are appropriate in certain
situations. The --auto flag tries to alleviate this problem and selects
accuracy/speed flags according to the number of sequences. For less
than 1,000 sequences mBed-clustering is unnecessary and the --mbed
flag is turned off. In this case the iteration is also turned off as
the effect of iteration is more noticeable for 'larger' problems. For
1,000 to 10,000 sequences full distance matrix calculation may become
to burdensome and the --mbed flag is turned on. Iterations are set to
1 as experience has shown that accuracy can be boosted in this
regime. However, for more than 10,000 sequences iterations are turned
off again. Improvement in accuracy may be substantial, however, the
computational overhead might be too costly for average users. In this
regime the --mbed flag remains turned on as full distance matrix
calculation may become prohibitive. Expert users may want to avoid
this flag and exercise more fine tuned control by selecting the
appropriate options manually.

Certain parts of the MSA calculation have been parallelised. Most
noticeably, the distance matrix calculation, and certain aspects of
the HMM building stage. Clustal-Omega uses OpenMP. By default,
Clustal-Omega will attempt to use as many threads as possible. For
example, on a 4-core machine Clustal-Omega will attempt to use 4
threads. The number of threads can be limited by setting the --threads
flag. This may be desirable, for example, in the case of
benchmarking/timing.

Help is available by specifying the -h flag.

By default Clustal-Omega does not print any information to stdout
(other than the final alignment, if no output file is
specified). Information concerning the progress of the alignment can
be obtained by specifying one verbosity flag (-v). This may be
desirable, to verify what Clustal-Omega is actually doing at the
moment. If two verbosity flags (-v -v) are specified, command-line
flags (explicitly and implicitly set) are printed in addition to the
progress report.  Triple verbose level (-v -v -v) is the most verbose
level. In addition to single- and double-verbose information much more
information is displayed: input sequences and names, details of the
tree construction and intermediate alignments. Tree construction
information includes pairwise distances. The number of pairwise
distances scales with the square of the number of sequences, and
double verbose mode is probably only useful for a small number of
sequences.

The current version number of Clustal-Omega can be displayed by
setting the --version flag.

The current version number of Clustal-Omega as well as the code-name
and the build date can be displayed by setting the --long-version
flag.

By default, Clustal-Omega does not over-write files. These can be (i)
alignment output, (ii) distance matrix and (iii) guide
tree. Overwriting can be forced by setting the --force flag.


EXAMPLES:

./clustalo -i globin.fa

Clustal-Omega reads the sequence file globin.fa, aligns the sequences
and prints the result to screen in fasta/a2m format.


./clustalo -i globin.fa -o globin.sto --outfmt=st

If the file globin.sto does not exist, then Clustal-Omega reads the
sequence file globin.fa, aligns the sequences and prints the result to
globin.sto in Stockholm format. If the file globin.sto does exist
already, then Clustal-Omega terminates the alignment process before
reading globin.fa.


./clustalo -i globin.fa	-o globin.aln --outfmt=clu --force

Clustal-Omega reads the sequence file globin.fa, aligns the sequences
and prints the result to globin.aln in Clustal format, overwriting the
file globin.aln, if it already exists.


./clustalo -i globin.fa --distmat-out=globin.mat --guidetree-out=globin.dnd --force

Clustal-Omega reads the sequence file globin.fa, aligns the sequences,
prints the result to screen in fasta/a2m format (default), the guide
tree to globin.dnd and the distance matrix to globin.mat, overwriting
those files if they already exist.


./clustalo -i globin.fa	--guidetree-in=globin.dnd

Clustal-Omega reads the files globin.fa and globin.dnd, skipping
distance calculation and guide tree creation, using instead the guide
tree specified in globin.dnd.


./clustalo -i globin.fa --hmm-in=PF00042.hmm

Clustal-Omega reads the sequence file globin.fa and the HMM file
PF00042.hmm (in HMMer2 or HMMer3 format).  It then performs the
alignment, transferring pseudo-count information contained in
PF00042.hmm to the sequences/profiles during the MSA.


./clustalo -i globin.sto

Clustal-Omega reads the file globin.sto (of aligned sequences in
Stockholm format). It converts the alignment into a HMM, de-aligns the
sequences and re-aligns them, transferring pseudo-count information to
the sequences/profiles during the MSA. The guide tree is constructed
using a full distance matrix of Kimura distances.


./clustalo -i globin.sto  --dealign

Clustal-Omega reads the file globin.sto (of aligned sequences in
Stockholm format). It de-aligns the sequences and then re-aligns
them. No HMM is produced in the process, no pseudo-count information
is transferred. Consequently, the output must be the same as for
unaligned output (like in the first example ./clustalo -i globin.fa)


./clustalo -i globin.fa --iter=2 

Clustal-Omega reads the file globin.fa, creates a UPGMA guide tree
built from k-tuple distances, and performs an initial alignment. This
initial alignment is converted into a HMM and a new guide tree is
built from the Kimura distances of the initial alignment. The
un-aligned sequences are then aligned (for the second time but this
time) using pseudo-count information from the HMM created after the
initial alignment (and using the new guide tree). This second
alignment is then again converted into a HMM and a new guide tree is
constructed. The un-aligned sequences are then aligned (for a third
time), again using pseudo-count information of the HMM from the
previous step and the most recent guide tree. The final alignment is
written to screen.


./clustalo -i globin.fa --iter=5 --max-guidetree-iterations=1

Clustal-Omega reads the file globin.fa, creates a UPGMA guide tree
built from k-tuple distances, and performs an initial alignment. This
initial alignment is converted into a HMM and a new guide tree is
built from the Kimura distances of the initial alignment. The
un-aligned sequences are then aligned (for the second time but this
time) using pseudo-count information from the HMM created after the
initial alignment (and using the new guide tree). For the last 4
iterations the guide tree is left unchanged and only HMM iteration is
performed. This means that intermediate alignments are converted to
HMMs, and these intermediate HMMs are used to guide the MSA during
subsequent iteration stages.


./clustalo -i globin.fa -o globin.a2m -v

In case the file globin.a2m does not exist, Clustal-Omega reads the
file globin.fa, prints a progress report to screen and writes the
alignment in (default) Fasta format to globin.a2m. The progress report
consists of the number of threads used, the number of sequences read,
the current progress in the k-tuple distance calculation, completion
of the guide tree computation and current progress of the MSA stage.
If the file globin.a2m already exists Clustal-Omega aborts before
reading the file globin.fa. Note that in verbose mode an output file
has to be specified, because progress/debugging information, which is
printed to screen, would interfere with the alignment being printed to
screen.


./clustalo -i PF00042_full.fa --dealign --mbed --outfmt=vie -o PF00042_full.vie --force 

Clustal-Omega reads the file PF00042_full.fa. This file contains
several thousand aligned sequences. --dealign tells Clustal-Omega to
erase all alignment information and re-align the sequences from
scratch. As there are several thousand sequences calculating a full
distance matrix may be slow. It may therefore be desirable to use the
--mbed option. Clustal-Omega now will calculate pairwise distances to
a small number of reference sequences only. This will give a
significant speed-up. The speed-up is greater for larger families
(more sequences). The alignment is then written out in Vienna format
(fasta format all on one line, no line breaks per sequence) to file
PF00042_full.vie.


./clustalo --p1=globin.sto --p2=PF00042_full.vie -o globin+pf00042.fa

Clustal-Omega reads files globin.sto and PF00042_full.vie of aligned
sequences (profiles). Both profiles are then aligned. The relative
positions of residues in both profiles are not changed during this
alignment, however, columns of gaps may be inserted into the profiles,
respectively. The final alignment is written to file globin+pf00042.fa
in fasta format.


./clustalo -i globin.fa --p1=PF00042_full.vie -o pf00042+globin.fa

Clustal-Omega reads file globin.fa of un-aligned sequences and the
profile (of aligned sequences) in file PF00042_full.vie. A HMM is
created from the profile. This HMM is used to guide the alignment of
the un-aligned sequences in globin.fa. The profile that was generated
during this alignment of un-aligned globin.fa sequences is then
aligned to the input profile PF00042_full.vie. The relative positions
of residues in profile PF00042_full.vie is not changed during this
alignment, however, columns of gaps may be inserted into the
profile. The final alignment is output to file pf00042+globin.fa in
fasta format. The alignment in this example may be slightly different
from the alignment in the previous example, because no HMM guidance
was used generate the profile globin.sto. In this example HMM guidance
was used to align the sequences in globin.fa; the hope being that this
intermediate alignment will have profited from the bigger profile.



LITERATURE:

[1] Johannes Soding (2005) Protein homology detection by HMM-HMM
    comparison. Bioinformatics 21 (7): 951–960.

[2] Blackshields G, Sievers F, Shi W, Wilm A, Higgins DG.  Sequence
    embedding for fast construction of guide trees for multiple
    sequence alignment.  Algorithms Mol Biol. 2010 May 14;5:21.

[3] http://www.genetics.wustl.edu/eddy/software/#squid

[4] Wilbur and Lipman, 1983; PMID 6572363

[5] Thompson JD, Higgins DG, Gibson TJ.  (1994). CLUSTAL W: improving
    the sensitivity of progressive multiple sequence alignment through
    sequence weighting, position-specific gap penalties and weight
    matrix choice. Nucleic Acids Res., 22, 4673-4680.

[6] Larkin MA, Blackshields G, Brown NP, Chenna R, McGettigan PA,
    McWilliam H, Valentin F, Wallace IM, Wilm A, Lopez R, Thompson JD,
    Gibson TJ, Higgins DG.  (2007). Clustal W and Clustal X version
    2.0. Bioinformatics, 23, 2947-2948.

[7] Kimura M (1980). "A simple method for estimating evolutionary
    rates of base substitutions through comparative studies of
    nucleotide sequences". Journal of Molecular Evolution 16: 111–120.

[8] Edgar, R.C. (2004) MUSCLE: multiple sequence alignment with high
    accuracy and high throughput.Nucleic Acids Res. 32(5):1792-1797.