Adaptive pruning option for local assembly #5473
Conversation
Codecov Report
@@ Coverage Diff @@
## master #5473 +/- ##
==============================================
+ Coverage 86.982% 87.07% +0.088%
- Complexity 31186 31244 +58
==============================================
Files 1914 1922 +8
Lines 144117 144210 +93
Branches 15933 15916 -17
==============================================
+ Hits 125356 125564 +208
+ Misses 13006 12872 -134
- Partials 5755 5774 +19
|
|
@davidbenjamin Thank you for doing this! Can you share the results you got on the mixtures? I'd be happy to try out this branch on our technical replicates next week. |
|
@davidbenjamin Fantastic! Have you talked to Sarah yet? Should I pass along a jar? |
|
@ldgauthier I gave her one about ten days ago. It looks fine so far on her RNA data. |
|
@meganshand I ran the "Full Pipeline" workflows in a clone of your FC workspace: https://portal.firecloud.org/#workspaces/broad-firecloud-dsde/copy-of-megans-m2-mito-validations. I did not run any of the things that generate graphs because they were harder for me to understand. To compare the new results to your previous ones, I took all variants that were either PASS or had only the contamination filter applied, extracted just the locus and alleles columns, then manually inspected the diff. For the 5% and 50% spike-ins there were usually no differences at all, while for the 1% spike-in the difference was usually 2-5 variants that straddled the LOD threshold. |
|
Great, thanks!
…On Mon, Dec 3, 2018 at 10:15 AM David Benjamin ***@***.***> wrote:
@meganshand <https://github.com/meganshand> I ran the "Full Pipeline"
workflows in a clone of your FC workspace:
https://portal.firecloud.org/#workspaces/broad-firecloud-dsde/copy-of-megans-m2-mito-validations.
I did not run any of the things that generate graphs because they were
harder for me to understand. To compare the new results to your previous
ones, I took all variants that were either PASS or had only the
contamination filter applied, extracted just the locus and alleles columns,
then manually inspected the diff. For the 5% and 50% spike-ins there were
usually no differences at all, while for the 1% spike-in the difference was
usually 2-5 variants that straddled the LOD threshold.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#5473 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AGRhdMwCcuQyzMweZjxWrXBODTCBaOSIks5u1T_-gaJpZM4Y9STI>
.
--
Laura Doyle Gauthier, Ph.D.
Associate Director, Germline Methods
Data Sciences Platform
[email protected]
Broad Institute of MIT & Harvard
320 Charles St.
Cambridge MA 0214
|
davidbenjamin
force-pushed
the
db_pruning
branch
from
a11aa05 to
bb66388
Compare
davidbenjamin
force-pushed
the
db_pruning
branch
from
bb66388 to
56849e1
Compare
davidbenjamin
force-pushed
the
db_pruning
branch
from
56849e1 to
ce8051c
Compare
|
@takutosato Based on all of our validations I added a commit to make this the default for M2. Because M2 shares a nested argument collection with HaplotypeCaller, this was pretty awkward. Louis told me this was the best among bad options. |
|
|
||
| import java.util.*; | ||
| import java.util.stream.Collectors; | ||
| import java.util.stream.IntStream; |
There was a problem hiding this comment.
some of these imports are not used
| return FastMath.max(leftLogOdds, rightLogOdds); | ||
| } | ||
|
|
||
| // is the chain |
There was a problem hiding this comment.
either remove or add a doc
|
|
||
| import static org.testng.Assert.*; | ||
|
|
||
| public class AdaptiveChainPrunerUnitTest { |
There was a problem hiding this comment.
Either remove or add tests here
There was a problem hiding this comment.
deleted -- the tests are all in ChainPRunerUnitTest.
|
back to @takutosato |
|
Oh wait, approving review, got it. |
Closes #4867.
@takutosato Here it is. I'm not quite ready to make it the M2 default, but it looks really good.
@meganshand I have tested it on every mixture in your workspace and results look very similar to the previous hand-tuned pruning results. I'm hoping it's good enough to become best practices for mitochondria and would appreciate if you gave it a shot. You have the right to review if you wish but there's no pressure to do so.
@ldgauthier HaplotypeCaller might also benefit from this. In particular, I wonder about #3697. I'll test it out.