Page SnippetsAnneKatrin

This is where I write my weekly goals.

all time todos:

• make splitRazers project page
• add link in seqanswers forum srp thread
• check sensitivity at 4x coverage and compare with dindel

6.12. - 10.12.

PLANNED

• Monday: write split mapping + SNP/indel calling part in Mat&Meth section. go through comments regarding miscalls/missing indels.

look at simulation pipeline results, complete runs

• Tuesday: include supersplat, include a generic indel detector
• Wednesday: write, write, plot, write
• Thursday: real results! make plan! --> Stefan (show first draft to him)
• Friday: send draft to Knut, should contain: methods, simresults, + a bit of intro

REALIZED:

• Monday: JC, mensa, commented comments, read dindel paper! --> they observe homopolymer-indels! and otherwise very low indel rates --> adapted simulation: mismatch probs on default, pi=pd=0.0002 (which still seems relatively high), started on queue. wrote half of MM text while falling asleep
• Tuesday: finished MM text. cluster down → restart simulation pipeline asap. took closer look at supersplat → cant do alignment errors + output is super raw, a lot of processing would have to be done --> raus!

29.11. - 3.12.

PLANNED

• Monday: adapt indelSimulator to simulate known indels within ranges (garuanteeing a certain number of indels per size range), set up smaller pipeline for testing purposes
• Tuesday: replace readsim with mason, include gsnap, bwa into pipeline
• Wednesday: include supersplat in pipeline, read up on other SE indel detection software
• Thursday: produce some example results, indelPercThresh 0/0.5, include 454 in pipeline
• Friday: meeting with Knut at 10.30, make new plan

REALIZED (or what i did instead..)

• Monday:
  • resolved snpStore issues for Stefan → input file was corrupt
  • discovered another pitfall for indel calling and realignment: read clipping may loose indels when clipped read is pairwise aligned to ref → minNumIndels required for realignment may not be reached → no indel call (solution: make indelThreshold for realignment independent from indelThreshold for indel calling + keep indel info for clipped reads (todo: friday))
  • fixed addInterval
  • read about indel rates: should be about 8-10:1 to snp rates
  • experimented with mason
• Tuesday:
  • adapted indelSimulator
  • replaced readsim with mason
• Wednesday:
  • set up small test pipeline, included gsnap, bwa
  • tested.. cant find sensible bwa settings
• Thursday:
  • added 454 simulation to pipeline
• Friday:

See progress report for the last 6 months.. all big todos are done!

big todos:

• indel calling on split-aligned reads in SeqAn
• gapped prefix/suffix alignment in razersSpliced
• realignment in snp calling

medium todos:

• generalized gff-to-fragmentStore parsing

small todos:

• open exchange calender (IE from home)

25.5. - 28.5.

PLANNED

• Tuesday: continued debugging
• Wednesday: ---
• Thursday: debugging/testing/code-cleaning
• Friday: 10.30 Kerstin

do also:

• make a more robust indel calling program --> collect positions in perl script or make c program!!!

17.5. - 21.5.

PLANNED

• Monday: make presentation, write review, check index construction in uniqueReads.cpp (done, sent to David)
• Tuesday: JC talk
• Wednesday: check Cougar's indels, sent improved microRazerS to Ho-Ryun, make split mapping run again
• Thursday: Kerstin 10.30 (vorher CNV-HMM angucken!), answer cougar,
• Friday: wrote tests for split alignment (still buggy obviously...)

3.5. - 7.5.

PLANNED

• process 454 data (doing!)
• debug split alignment for edit distance
• make plan for split-mapping results!!!!!!!!
• answer kerstin, sabrina, lena
• continue writing paper

26.4. - 30.4.

PLANNED

• process 454 data (doing!)
• finish split alignment for edit distance (done!)
• make plan for split-mapping results!!!!!!!!

kleines todo fuer dienstag abend: splitmapping auf A14 unmapped batches anschmeissen!!! sabrina antworten, lena antworten (A14 mappings, snps, indels, + readme! + mapped split? )

19.4. - 23.4.

PLANNED

• process 454 data
• finish split alignment for edit distance!!!
• make plan for split-mapping results for ismb/paper?

REALIZED

• split edit debugging....

12.4. - 16.4.

PLANNED

• process 454 data
• finish split alignment for edit distance!!!
• make plan for split-mapping results for ismb/paper?
• read and take notes about other split-mappers
• look into realignment again...

monday: send travel fellowship application, put progress report online, answer lena, finish edit split mapping!

5.4. - 8.4.

PLANNED

• progress report
• process 454 data
• split alignment for edit distance!

22.3. - 26.3.

PLANNED

• split alignment for edit distance! --> not done yet
• reprocess some patients with new split alignment (done)
• meet Kerstin and maybe try simple CNV detection on our data (not ready for testing yet)
• 25. & 26. Lena aus Tübingen

15.3. - 19.3.

PLANNED

• write poster abstract
• resolve pindel calling problems
• finish SeqAn tutorial

REALIZED

• sent abstract to ISMB
• resolved split alignment pindel calling problems
• integrated random match estimate into split alignment
• tutorial as good as finished

8.3. - 12.3.

PLANNED

• make plan for poster!
• breakpoint statistics: adapt parameters and check influence of windowlength, make plots
• continue SeqAn tutorial

REALIZED

• expected number of random matches is extremely low for 76bp reads with 23bp minMatchLen, even with errors, and even on whole genome scale → iid model probably not good enough to approximate real-genome-situation

1.3. - 5.3.

PLANNED

• fix split-alignment (done) * missing: maskDuplicates for spliced matches (done for hamming distance)
• fix edit-distance-indel-calling bug(XLMR_1996, siehe Stefans mail) (done)
• continue SeqAn tutorial (continuing)
• get snpStore ready for Marcel (done)
• check out Marcel's breakpoint statistics, adapt parameters and check influence of windowlength (halfway done)

REALIZED

• finished snpStore for Marcel, met on Tuesday
• fixed edit-distance-indel-calling bug: wrong indel position was calculated for reverse gapped reads
• seemingly fixed split-alignment, instead of maskDuplicates: directly discard duplicate prefix and duplicate suffix matches (only works for hamming distance)

22.2. - 26.2.

PLANNED

• test split-alignment
• implement indel calling on split-aligned reads
• continue reading and taking notes
• integrate indel calling on realigned reads
• meet with Paz
• do realignment on windows of varying size? inspect influence of window size. debug realigner?
• SeqAn tutorial

REALIZED

• split-alignment --> model gap costs with exponential funktion? → talked to Marcel about problem of read placement, + probabilities of random matches, expected numbers of reads → variant predicition
• somehow produced a weird bug in split-alignment when integrating into uptodateSeqan
• SeqAn tutorial: finished motif finding + started with alignments
• met with Kerstin --> seqan basics, structs, fragmentStore
• seminars: reseq meeting, genereg meeting, group meeting, Illumina Casava 1.6 talk by Oliver Goldenberg

15.2. - 19.2.

PLANNED

• test split-alignment --> weird insertion/deletion calls (doing)
• implement indel calling on split-aligned reads (not done)
• continue reading and taking notes (naja)
• integrate indel calling on realigned reads (not done)
• meet with Paz (next week)
• do realignment on windows of varying size? inspect influence of window size. debug realigner? (not done)
• meet with Marcel Grunert (done)
• meet with Bernd Timmermann --> casava 1.6, heterogeneous dataset (done)

REALIZED

• test split-alignment --> weird insertion/deletion calls: problem in sorting of matches which was done like this 1) minimizing number of errors (not even minimizing of distanceError was implemented!!!). instead
• met with Marcel Grunert: need to adapt snpStore to incorporate readcounts given as gff-tag. do snp calling on mircoRNAs.
• met with Bernd Timmermann

1.2. - 5.2.

PLANNED

• fix razers uniqueness bug (dachte dass done, aber doch nicht...)
• correct maxPile correction for indel calls (done)
• integrate indel calling on realigned reads (not done)
• meet with Paz (not done)
• do realignment on windows of varying size? inspect influence of window size. debug realigner? (not done)
• continue reading and taking notes (doing)

REALIZED

• Monday: meeting with Kerstin, discussed different CNV tools
• Tuesday: fixed razers uniqueness bug in all three versions
• Wednesday: checked distribution of call quality scores and presented in reseq meeting
• Thursday: integrated indel calling, now insertion/deletion seperate in gff
• Friday: tested snpStore and razerS

25.1. - 29.1.

PLANNED

major:

• test/debug snp calling on realigned reads (doing)
• read SV detection papers and take notes in wiki (doing now)

minor:

• answer tübingen people (done)
• read MA Vipul (done)

REALIZED

• Realignment: problem is in including the reference sequence into the multi read alignment. If includeReference == true this only means that the original reference sequence is stored as an artificial read that is NOT realigned. if one includes the artificial reference-read in the realignment process, the reference is not given enough weight, ie many artificial snp calls may be the result. instead one needs to do a pairwise alignment of reference and consensus sequence, to get a better reference-to-reads alignment.

--> Did that. Results are worse. TODO: inspect cases to find cause. maybe a problem with static windowlength?

• Inspected quality scores of SNP calls: the more FP calls, the more their distribution resembles a geometric distribution. GC bias is also detected in high-quality calls, even gets stronger with increasing quality for geometrically distributed quality scores. --> base call quality value recalibration needed? (to correct for nucleotide specific bias, downweight reads with many errors, ie assign mapping quality) more stringent clipping? (stefan)

Comments