SeqIO dev

From Biopython
Revision as of 14:09, 2 July 2008 by Maubp (Talk | contribs)
Jump to: navigation, search

This page is aimed at any developers or coders interesting in understanding or extending the new Sequence Input/Output interface for BioPython, SeqIO.

The code has now been checked into CVS. Related Bug 2059 has been resolved.

The code is already available in BioPython 1.43.

Contents

Reading new file formats

Note: The details are still subject to change

To add support for reading a new file format, you must implement an iterator that expects a just file handle and returns SeqRecord objects. You may do this using:

  • An iterator class subclassing something from Bio.SeqIO.Interfaces
  • A generator function (using the yield keyword; suitable for simple formats)
  • An ordinary function which returns an iterator. For example, you could build a list of SeqRecords and then turn it into an iterator using the iter() function.

You may accept additional optional arguments (an alphabet for example). However there must be one and only one required argument (the input file handle).

What you use as the SeqRecord's id, name and description will depend on the file format. Ideally you would use the accesion number for the id. This id should also be unique for each record (unless the records in the file are in themselves ambiguous).

When storing any annotations in the record's annotations dictionary follow the defacto standard laid down by the GenBank parser... I should try and document this more.

If the supplied file seems to be invalid, raise a ValueError exception.

Finally, the new format must be added to the relevant dictionary mapping in Bio/SeqIO/__init__.py so that the Bio.SeqIO.parse() and Bio.SeqIO.read() functions are aware of it.

Writing new file formats

Note: The details are still subject to change

To add support for writing a new file format you should write a sub class of one of the writer objects in Bio.SeqIO.Interfaces

Then, the new format must be added to the relevant dictionary mappings in Bio/SeqIO/__init__.py so that the Bio.SeqIO.write() function is aware of your code.

If the supplied records cannot be written to this file format, raise a ValueError exception. Where appropriate, please use the following wording:

raise ValueError("Must have at least one sequence")
raise ValueError("Sequences must all be the same length")
raise ValueError("Duplicate record identifier: %s" % ...)
...

ToDo - Defined standard exceptions in Bio.SeqIO itself?

Possible additional formats

There are existing parsers in BioPython for the following file formats, which could be integrated into Bio.SeqIO if appropriate.

NBRF / PIR format

Bio.NBRF has a Martel parser for PIR sequence format, which is similar to the FASTA format. It would need addition work to return SeqRecords. It might be easier to extend to reuse the Bio.SeqIO fasta code instead.

There is also PSC documentation.

KEGG format

Can Bio.KEGG parse files in KEGG format?

IntelliGenetics / MASE alignment format

Bio.IntelliGenetics seems to use Martel parse MASE format like files into its own record object. It could be extended to return SeqRecord objects. There are already sample input files in Biopython for the unit tests, but these are clearly not multiple sequence alignments. See also this EMBOSS example:

http://emboss.sourceforge.net/docs/themes/seqformats/ig

And these pages about MASE files:

http://pbil.univ-lyon1.fr/help/formats.html http://www.bioperl.org/wiki/Mase_multiple_alignment_format

Biopython-dev mailing list discussion: http://lists.open-bio.org/pipermail/biopython-dev/2008-July/003933.html

MEME format

Bio.MEME has a parser for this file format, which at first glance looks like it could be treated like an alignment format.

http://meme.sdsc.edu

BLAST results

Pairwise alignments from the BLAST suite could be turned into two SeqRecord objects with gapped sequences. Is this useful?

COMPASS pairwise alignment format

Bio.Compass can parse the pairwise alignments from COMPASS. The output is similar to BLAST in many ways. Again, is getting the results as SeqRecord objects useful?

Personal tools
Namespaces
Variants
Actions
Navigation
Toolbox