[Biopython] Struggling with MSA...

[Peter Cock]

You may have more joy starting from the HMM for the model?

https://www.ebi.ac.uk/interpro/wwwapi//entry/pfam/PF08241?annotation=hmm

Peter

On Thu, Feb 15, 2024 at 4:06 PM Dan Bolser <dan.bolser at outsee.co.uk> wrote:
>
> Searching HMMER gives me (cryptic?) details of the model match:
>
> Family-Id Family-Accession Clan Start End Ali-Start Ali-End Model-Start Model-End Bit-Score Ind.-E-value Cond.-E-value Description
> ==================================================================================================================================
>
> Methyltransf_11 PF08241.15 CL0063 63 173 63 172 1 95 51.30 1.5e-13 3.1e-17 Methyltransferase domain
>
> ---------------------------------------------------------------------------------------------------------------------------------
>
> MODEL  LdvGcGtGrlaealakrg.arvvgvDlskemlklakekaseeglkvefvvadaeklpfednsfDlvvssevlhhv...e............dpekalkeiaRvLkpgGllv
> MATCH  L +GcG+ +l+ +l   g  +v+ vD+s  ++++ +++    + ++++  +d++kl f+++sfD+v+ + +l+ +   e             ++++l+e+ RvL pgG+++
> PPL    569************7779*************77766666666.69*****************************8662667778888888******************97
> SEQ    LVLGCGNSALSYELFLGGfPNVTSVDYSSVVVAAMQARHAHVP-QLRWETMDVRKLDFPSASFDVVLEKGTLDALlagErdpwtvssegvhTVDQVLSEVSRVLVPGGRFI
>
>
> So I guess it's possible to work backwards from there...
>
>
>
> On Thu, 15 Feb 2024 at 16:02, Peter Cock <p.j.a.cock at googlemail.com> wrote:
>>
>> Sorry yes, I used this as a threshold for if a column was gappy, and
>> likely to be in the model or not:
>>
>> amino_acids.get("-", 0) < len(align)*0.5
>>
>> Might need to use <= to match, or perhaps they use a more
>> sophisticated criteria.
>>
>> Peter
>>
>> On Thu, Feb 15, 2024 at 3:56 PM Dan Bolser <dan.bolser at outsee.co.uk> wrote:
>> >
>> > Sorry, by 'half gaps' I thought you meant a specific thing, like, "look at that half gap!". I guess you mean, columns where half of the values in the column are gaps.
>> >
>> > Looking directly at the alignment, I guess there is no way to work out precisely which columns are gaps or not in the model?
>> >
>> >
>> >
>> >
>> >
>> >
>> > On Thu, 15 Feb 2024 at 15:47, Dan Bolser <dan.bolser at outsee.co.uk> wrote:
>> >>
>> >> Hi Peter,
>> >>
>> >> On Thu, 15 Feb 2024 at 15:28, Peter Cock <p.j.a.cock at googlemail.com> wrote:
>> >>>
>> >>> Hello Dan,
>> >>>
>> >>> It that is a probability at the end, you have the wrong denominator -
>> >>> should be len(align) == sum(amino_acids.values()
>> >>
>> >>
>> >> Makes no difference, they are the same (including -s). The profile reports -s as 'Insert Probability'.
>> >>
>> >>
>> >>> Note this seed alignment has 164 columns, yet the model page says the
>> >>> model has 95 columns. If I count the number of columns which are half
>> >>> gaps that's pretty close...
>> >>
>> >>
>> >> I don't understand what that means or implies... Is something wrong somewhere?
>> >>
>> >>
>> >> Cheers,
>> >> Dan.
>> >>
>> >>
>> >>> Peter
>> >>>
>> >>> On Thu, Feb 15, 2024 at 3:09 PM Dan Bolser <dan.bolser at outsee.co.uk> wrote:
>> >>> >
>> >>> > Hi,
>> >>> >
>> >>> > Sorry, I can't follow the docs (or find the right docs).
>> >>> >
>> >>> > I've got the 'seed' stockholm alignment for this domain:
>> >>> > https://www.ebi.ac.uk/interpro/entry/pfam/PF08241/entry_alignments/?type=seed
>> >>> >
>> >>> > and I'm trying to reproduce the signature it shows here:
>> >>> > https://www.ebi.ac.uk/interpro/entry/pfam/PF08241/logo/
>> >>> >
>> >>> > I'm not sure a) why the probabilities differ in the profile relative to the seed alignment, or b) how to filter columns in the alignment by those that have a match in the model (see columns 4-6 in the alignment, which are gaps in the model).
>> >>> >
>> >>> > I think if I can answer b) then the answer to a) will be, "look at the full alignment".
>> >>> >
>> >>> > Here is my crude 'best guess' code:
>> >>> >
>> >>> > import gzip
>> >>> > import Bio.AlignIO
>> >>> >
>> >>> > # msa = "PF08241.alignment.full.gz"
>> >>> > msa = "PF08241.alignment.seed.gz"
>> >>> >
>> >>> > with gzip.open(msa, "rt") as handle:
>> >>> > align = Bio.AlignIO.read(handle, "stockholm")
>> >>> > ncols = align.get_alignment_length()
>> >>> >
>> >>> > for col in range(ncols):
>> >>> > amino_acids = dict()
>> >>> > for s in align[:, col]:
>> >>> > amino_acids[s] = amino_acids.get(s, 0) + 1
>> >>> > print(amino_acids)
>> >>> > for s in amino_acids:
>> >>> > print(f"{s}: {amino_acids[s]:3d} {amino_acids[s] / len(align):.3f}")
>> >>> >
>> >>> >
>> >>> >
>> >>> > I have the feeling I'm doin it rong...
>> >>> >
>> >>> > The above is just a 'warm up', really I want to see the conservation score, base by base on a given protein in the alignment (where it matches the model).
>> >>> >
>> >>> > Many thanks for any suggestions, and sorry for not being able to find the right document to answer these questions.
>> >>> >
>> >>> >
>> >>> > kthxbi,
>> >>> > Dan.
>> >>> > _______________________________________________
>> >>> > Biopython mailing list  -  Biopython at biopython.org
>> >>> > https://mailman.open-bio.org/mailman/listinfo/biopython