Aug. 21-26, 2019 Geoduck genome paper-DMRs

Aug. 21 Conference call

paper repo:

• Decided on Hollie’s repo: Geoduck_Meth

Current v074 .gff:

• 8400 gene list that correspond to 5000 proteins
• TE gff on genomic resources on github

DMG analysis:

• doesn’t require same loci; just % methylation across DMGs
• finds common gene loci that tend to be more methylated
• Comparisons: see supp tables in manuscript

DML analysis:

• Steven is working on DMLs to compare loci across bivalves

Strandedness:

• After running Mac’s code, Hollie’s output goes from 8M in bam to 6M in merge.cov
  • code deletes bottom strand coverage of CG
• We decided to destrand in all our analyses

Coverage:

• We decided to use 5x and 10x, and will see which is best after analysis

Global analysis of genome methylation

Describe where is methylation in genome:

• Steven will run bismark on concatenated data
  • while this was running I generated a code to concatenate .cov.gz files from Bismark coverage analysis
    • practice code here: https://github.com/shellywanamaker/Shelly_Pgenerosa/blob/master/analyses/JuviPgen_ALSL2lowd145/Concatenate_Covgz_files.ipynb
    • final code here: https://gannet.fish.washington.edu/metacarcinus/mox_jobs/20190821_concatCovgz.sh
    • output here: https://gannet.fish.washington.edu/metacarcinus/Pgenerosa/analyses/20190821/uber_methylome.cov.gz
• With one coverage file:
  • bedtools intersect on genome features (.gff) and .cov.gz file
  • for each feature, sum CpG?
  • sashimi plots?

DMR calling

Methylpy: Destrand data

• I didn’t find an option to destrand the data so I created a code to destrand the allc files and then run DMRfind on the destranded allc files: https://github.com/shellywanamaker/Shelly_Pgenerosa/blob/master/analyses/JuviPgen_ALSL2lowd145/Get_DMRs_for_SR_v074_alignments/methylpy_DMRs/destrand_allc_files/Destrand_allc_files.ipynb
• this was before Yupeng sent me his code for destranding
• I did find 10 more DMRs after destranding using the same DMRfind settings: 56 DMR vs. 46 DMRs (see table below)

Methylkit: adjust stringency

• I had previously set the methylkit unite function to select common sites in samples if they were in >= 2 samples (1L)

• I tried 3L to see if that would give less false positive DMRs

• I also tried the default min.group parameter (requires all samples to show coverage of loci)

*Methylkit R project here: https://github.com/shellywanamaker/Shelly_Pgenerosa/blob/master/analyses/JuviPgen_ALSL2lowd145/Get_DMRs_for_SR_v074_alignments/Get_DMRs_for_SR_v074_alignments.Rproj

IGV Sessions

• session comparing DMRs from different Methylkit parameters

• session visualizing methylpy DMRs NOT-destranded, 3x coverage of 1000bp windows containing minimum of 3 DMLs (DML = 25bp chunk): https://github.com/shellywanamaker/Shelly_Pgenerosa/blob/master/analyses/JuviPgen_ALSL2lowd145/20190819_Pgnrv074_DMRs_in_IGV/20190819_Pgnrv074_DMRs_in_IGV.xml
  • jupyter notebook for preparing files and loading them into IGV: https://github.com/shellywanamaker/Shelly_Pgenerosa/blob/master/analyses/JuviPgen_ALSL2lowd145/20190819_Pgnrv074_DMRs_in_IGV/20190819_Pgnrv074_DMRs_in_IGV.ipynb
• session comparing methylpy DMRs destranded versus stranded: https://github.com/shellywanamaker/Shelly_Pgenerosa/blob/master/analyses/JuviPgen_ALSL2lowd145/Get_DMRs_for_SR_v074_alignments/methylpy_DMRs/destrand_allc_files/destranded_v_stranded.xml

Some DMRs don’t get called after destranding, even though the DMSs are still there. Why is this? Example = scaffold 6 42570880-42570905