Why didn't we find this variant supported?

[ihodes]

This variant didn't make it into the report for some reason (this is RNA support being shown), all well-aligned good data.

[ihodes]

Also one like this

[iskandr]

These are all by exon boundaries (I think -- can't see the annotation track in these screen shots). After dropping all the multi-exon spliced reads we're left with a bunch of reads that contain intronic sequences (blame RiboZero). Since the intronic sequences disagree with all annotated transcripts then we never end up discovering a reading frame for these variants.

[ihodes]

Would it be reasonable and feasible to stop reading cDNA at exon boundaries, so when we inevitably get intronic RNA in some reads, they aren't effectively thrown out?

[iskandr]

We can certainly rely on the reference more, particularly if we give up trying to detect altered splicing.

…

On Dec 11, 2016 10:27 AM, "Isaac Hodes" ***@***.***> wrote: Would it be reasonable and feasible to stop reading cDNA at exon boundaries, so when we inevitably get intronic RNA in some reads, they aren't effectively thrown out? — You are receiving this because you were assigned. Reply to this email directly, view it on GitHub <#97 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAC9OUwyrBMVuWQKClMrz_BDf6l1OcI8ks5rG8HzgaJpZM4LID9j> .

[ihodes]

Ah interesting, where/how are we trying to detect altered splicing? On Sun, Dec 11, 2016 at 9:19 AM, Alex Rubinsteyn <[email protected]> wrote:

…

We can certainly rely on the reference more, particularly if we give up trying to detect altered splicing. On Dec 11, 2016 10:27 AM, "Isaac Hodes" ***@***.***> wrote: > Would it be reasonable and feasible to stop reading cDNA at exon > boundaries, so when we inevitably get intronic RNA in some reads, they > aren't effectively thrown out? > > — > You are receiving this because you were assigned. > Reply to this email directly, view it on GitHub > <#97 (comment)>, > or mute the thread > <https://github.com/notifications/unsubscribe- auth/AAC9OUwyrBMVuWQKClMrz_BDf6l1OcI8ks5rG8HzgaJpZM4LID9j> > . > — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#97 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAKPe3_ITPq2GxQztw2qDvbIvc8swvDxks5rHAZugaJpZM4LID9j> .

[ihodes]

@julia326 @iskandr can we close this/have we solved this?

[julia326]

I don't think we detect altered splicing (@iskandr please chime in if I'm wrong about this) - and the recent change to count intronic reads as reference mismatches after the exon boundary will actually make us throw out more reads. Coupled with these Ribozero side effects, that could be problematic. Is this a variant we would've wanted to keep, or should we double down on working with poly-A capture in the future?

[iskandr]

@julia326 Define "detect" :-) -- we were certainly detecting it in this case, just incorrectly.

[iskandr]

@ihodes How do these variants look after our changes to indel realignment?

[iskandr]

@ihodes @julia326 When samples are degraded and/or prepared with ribosomal depletion we should count mismatches on both sides of the variant, which will penalize variant sequences that look like abnormal splicing.

[iskandr]

Closing, since this is really an isovar issue: openvax/isovar#55 (which will be fixed by openvax/isovar#80)