CLI

[iskandr]

Gives isovar a commandline interface and CSV output (#6)

[hammer]

I have some code review bandwidth and am interested in isovar. Does this PR represent the latest state of the project?

[iskandr]

Not quite, there are some fragments not checked in. I'll give you something
reviewable tomorrow though.
On Mar 30, 2016 1:06 AM, "Jeff Hammerbacher" notifications@github.com
wrote:

I have some code review bandwidth and am interested in isovar. Does this
PR represent the latest state of the project?

—
You are receiving this because you authored the thread.
Reply to this email directly or view it on GitHub
#7 (comment)

[hammer]

@iskandr ready for review?

[iskandr]

@hammer Not yet, still working through some edge cases I thought of last night. The parts that don't interact with read data (such as the reference_context module) are pretty stable though, if you want to start somewhere.

[iskandr]

@hammer variant_reads, variant_sequences, and reads_at_locus should also be more or less in their final shape, with passing unit tests. The only missing part now is that the translation logic is split across a few incomplete files, which I need to weave back together.

[iskandr]

@tavinathanson @ihodes @hammer If you all want to start reviewing, anything other than protein_sequence and its tests are fair game.

I added some explanation of the design to the README but it's worth repeating a little here:

1. Reads are filtered from the BAM around each variant into ReadAtLocus records. These contain the sequence of a read and the offsets into that sequence for the bases immediately before and after a variant. I found "anchoring" on reference bases adjacent to the variant was easier than trying to deal with the variant locus directly.

2. Not every ReadAtLocus object contains the variant nucleotides, so these get further filtered down into VariantRead objects.

3. Multiple VariantReads get collapsed into a VariantSequence. These are the cDNA sequences we're going to try translating.

4. Which reading frame to use? We're not really sure (without long reads + ribosomal footprint profiling), so instead we take a reasonable guess based on the reading frames of reference transcripts around this locus. I summarize the reference sequence at this locus and its associated reading frame in a ReferenceContext object. There's a weird little hierarchy of named tuples in that file, which I found simplified testing a lot (enabled me to right more tests and to think more clearly about little bits of functionality).

5. Finally, the VariantReads and ReferenceContexts of a variant get combined into zero or more ProteinSequence objects. Since we may only want the single best translation, both reference contexts and variant sequences are sorted first.

[tavinathanson]

Just poking around; noticed that you have two base1_position_before_variantss vs. after.

[coveralls]

Changes Unknown when pulling c8ee0bf on cli into * on master*.

[coveralls]

Changes Unknown when pulling 3959de7 on cli into * on master*.

[coveralls]

Changes Unknown when pulling 1706af7 on cli into * on master*.

[coveralls]

Changes Unknown when pulling 1706af7 on cli into * on master*.

[tavinathanson]

@iskandr I'm running into AttributeError: 'module' object has no attribute 'cutils' with pysam 0.8.4; doesn't complain with 0.9.0.

[coveralls]

Changes Unknown when pulling bdd3125 on cli into * on master*.

[coveralls]

Changes Unknown when pulling 4910ded on cli into * on master*.

[coveralls]

Changes Unknown when pulling d06c34e on cli into * on master*.

[coveralls]

Changes Unknown when pulling a38fe9d on cli into * on master*.

[iskandr]

Fixes #12, #11, #10, #6.

I need to more explicitly test for #4 since the B16 variants do have nearby germline variants but I haven't documented what they are.