Check on screed's attributes

[ctb]

Does 'accuracy' still exist on FASTA records?

#769 (comment)

Update: last remaining differences between khmer Read and screed Record objects seems to be description vs annotation.

[luizirber]

No, neither does quality.

• khmer.ReadParser records also use annotations, while screed.screedRecord uses description.
• khmer.Read has everything from the ID line in the name field, leaves annotations empty
• screed.screedRecord has ID line split by spaces, first into name and remainder in description

[camillescott]

Note that the quality attribute always seems to exist in the ReadParser's Read extension class, so the hasattr(read, "quality") paradigm doesn't work for them. Perhaps both implementations should switch to adding an extra format attribute.

[standage]

If .quality is always present but never set for Fasta records, then it should be possible to switch the check from hasattr(record, "quality") to record.quality != "", no?

[camillescott]

Yup, although screed just leaves it off entirely if it doesn't exist. I'd
prefer using a @property and returning None for the null case (checking
is None has less room for edge cases than empty string).

On Thu, Nov 17, 2016 at 9:50 AM, Daniel Standage notifications@github.com
wrote:

If .quality is always present but never set for Fasta records, then it
should be possible to switch the check from hasattr(record, "quality") to record.quality
!= "", no?

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
#1484 (comment), or mute
the thread
https://github.com/notifications/unsubscribe-auth/ACwxrbGPMKxN0A4rJ2I68uvDS4RsSTonks5q_JPYgaJpZM4KP8NG
.

Camille Scott

Graduate Group for Computer Science
Lab for Data Intensive Biology
University of California, Davis

camille.scott.w@gmail.com

[ctb]

This was discussed in #1468 and resolved to eliminating the quality attribute on FASTQ records, which was the opinion of me and @luizirber both. Following from that, we decided 'quality' should be eliminated from ReadParser results on FASTA sequences.

At this point I'm happy to reconsider (I kinda like the 'format' notion) but I would like a compleat (if brief) proposal that satisfies all of the discussion points raised in the various discussions, i.e. would require some work by someone :).

[ctb]

One extra consideration is whether Nanopore or PacBio add sequence attributes that we want to plan ahead for. I don't have a clear idea of this.

[betatim]

For me read = khmer.Read("blah", "ACGT"); hasattr(read, 'quality') works as expected (returns False). (However ipython tab-completes read.q to read.quality so somewhere there is a bug)

[ctb]

I believe #1484 (comment) is incorrect - 'description' is only set if parse_description=True.

[ctb]

But it seems like the bigger annoyance is that khmer now has 'annotations' and screed has 'description' (when parse_description is set)

[ctb]

Note that we must allow 'screed.Record()', as it is used in khmer 2.0 which depends on screed > 0.9. In future we should remember to bound requirements by next major version :)

[betatim]

See dib-lab/screed#64 for allowing Read() to work again.

[standage]

screed.Record() compatibility between khmer and screed was restored with dib-lab/screed#64. As far as bringing reads/records into alignment between khmer and screed, it looks like there are two remaining discrepancies.

• annotations attribute in khmer vs description attribute in screed (Check on screed's attributes #1484 (comment))
• quality attribute present in khmer.Read but not screed.Record for fasta records (screed.Record(..., quality=None) should not set 'quality' #1468)

[ctb]

quality does not appear to be present in khmer.Read(...) if quality=None in constructor.

See test_read_parsers.py::test_read_type_basic and test_read_parsers.py::test_read_quality_none (added in #1559).

One remaining question - is it present when we read FASTA sequences?

[betatim]

In [2]: rp = khmer.ReadParser('tests/test-data/random-20-a.fa')

In [3]: for read in rp:
   ...:     print(read)
   ...:     print(hasattr(read, 'quality'))
   ...:     
<khmer.Read object at 0x7f31b9e4acc0>
False

Shall we immortalise by making it a test?

[ctb]

See #1583.

The only remaining issue here seems to be 'annotations' which we should be able to change in khmer (since AFAIK nothing uses it...)

[betatim]

I'm renaming annotations to description.