Building alignments

Using the cogent aligners

Running a pairwise Needleman-Wunsch-Alignment

>>> from cogent.align.algorithm import nw_align
>>> seq1 = 'AKSAMITNY'
>>> seq2 = 'AKHSAMMIT'
>>> print nw_align(seq1,seq2)
('AK-SAM-ITNY', 'AKHSAMMIT--')

Running a progressive aligner

We import useful functions and then load the sequences to be aligned.

>>> from cogent import LoadSeqs, LoadTree, DNA
>>> seqs = LoadSeqs('data/test2.fasta', aligned=False, moltype=DNA)

For nucleotides

We load a canned nucleotide substitution model and the progressive aligner TreeAlign function.

>>> from cogent.evolve.models import HKY85
>>> from cogent.align.progressive import TreeAlign

We first align without providing a guide tree. The TreeAlign algorithm builds pairwise alignments and estimates the substitution model parameters and pairwise distances. The distances are used to build a neighbour joining tree and the median value of substitution model parameters are provided to the substitution model for the progressive alignment step.

>>> aln, tree = TreeAlign(HKY85(), seqs)
Param Estimate Summary Stats: kappa
==============================
        Statistic        Value
------------------------------
            Count           10
              Sum        1e+06
           Median        4.256
             Mean        1e+05
StandardDeviation    3.162e+05
         Variance        1e+11
------------------------------
>>> aln
5 x 60 text alignment: Mouse[GCAGTGAGCCA...], NineBande[-C-----GCCA...], Human[GCAAGGAGCCA...], ...

We then align using a guide tree (pre-estimated) and specifying the ratio of transitions to transversions (kappa).

>>> tree = LoadTree(treestring='(((NineBande:0.0128202449453,Mouse:0.184732725695):0.0289459522137,DogFaced:0.0456427810916):0.0271363715538,Human:0.0341320714654,HowlerMon:0.0188456837006)root;')
>>> params={'kappa': 4.0}
>>> aln, tree = TreeAlign(HKY85(), seqs, tree=tree, param_vals=params)
>>> aln
5 x 60 text alignment: NineBande[-C-----GCCA...], Mouse[GCAGTGAGCCA...], DogFaced[GCAAGGAGCCA...], ...

For codons

We load a canned codon substitution model and use a pre-defined tree and parameter estimates.

>>> from cogent.evolve.models import MG94HKY
>>> tree = LoadTree(treestring='((NineBande:0.0575781680031,Mouse:0.594704139406):0.078919659556,DogFaced:0.142151930069,(HowlerMon:0.0619991555435,Human:0.10343006422):0.0792423439112)')
>>> params={'kappa': 4.0, 'omega': 1.3}
>>> aln, tree = TreeAlign(MG94HKY(), seqs, tree=tree, param_vals=params)
>>> aln
5 x 60 text alignment: NineBande[------CGCCA...], Mouse[GCAGTGAGCCA...], DogFaced[GCAAGGAGCCA...], ...

Building alignments with 3rd-party apps such as muscle or clustalw

See Using alignment application controllers to align unaligned sequences.

Converting gaps from aa-seq alignment to nuc seq alignment

We load some unaligned DNA sequences and show their translation.

>>> from cogent import LoadSeqs, DNA, PROTEIN
>>> seqs = [('hum', 'AAGCAGATCCAGGAAAGCAGCGAGAATGGCAGCCTGGCCGCGCGCCAGGAGAGGCAGGCCCAGGTCAACCTCACT'),
...         ('mus', 'AAGCAGATCCAGGAGAGCGGCGAGAGCGGCAGCCTGGCCGCGCGGCAGGAGAGGCAGGCCCAAGTCAACCTCACG'),
...         ('rat', 'CTGAACAAGCAGCCACTTTCAAACAAGAAA')]
>>> unaligned_DNA = LoadSeqs(data=seqs, moltype = DNA, aligned = False)
>>> print unaligned_DNA.toFasta()
>hum
AAGCAGATCCAGGAAAGCAGCGAGAATGGCAGCCTGGCCGCGCGCCAGGAGAGGCAGGCCCAGGTCAACCTCACT
>mus
AAGCAGATCCAGGAGAGCGGCGAGAGCGGCAGCCTGGCCGCGCGGCAGGAGAGGCAGGCCCAAGTCAACCTCACG
>rat
CTGAACAAGCAGCCACTTTCAAACAAGAAA
>>> print unaligned_DNA.getTranslation()
>hum
KQIQESSENGSLAARQERQAQVNLT
>mus
KQIQESGESGSLAARQERQAQVNLT
>rat
LNKQPLSNKK

We load an alignment of these protein sequences.

>>> aligned_aa_seqs = [('hum', 'KQIQESSENGSLAARQERQAQVNLT'),
...                    ('mus', 'KQIQESGESGSLAARQERQAQVNLT'),
...                    ('rat', 'LNKQ------PLS---------NKK')]
>>> aligned_aa = LoadSeqs(data=aligned_aa_seqs, moltype=PROTEIN)

We then obtain an alignment of the DNA sequences from the alignment of their translation.

>>> aligned_DNA = aligned_aa.replaceSeqs(unaligned_DNA, aa_to_codon=True)
>>> print aligned_DNA
>hum
AAGCAGATCCAGGAAAGCAGCGAGAATGGCAGCCTGGCCGCGCGCCAGGAGAGGCAGGCCCAGGTCAACCTCACT
>mus
AAGCAGATCCAGGAGAGCGGCGAGAGCGGCAGCCTGGCCGCGCGGCAGGAGAGGCAGGCCCAAGTCAACCTCACG
>rat
CTGAACAAGCAG------------------CCACTTTCA---------------------------AACAAGAAA

Setting the argument aa_to_codons=False is only useful when the sequences have exactly the length. One use case is to allow introducing the gaps onto another copy of the alignment where there are annotations.