I basically followed exactly along with the methods from the Diagenode kit with some very minor adjustments so I’m not going to describe it here. I have a file with all the lab notes and sample details. Sequencing was at BRC and I have the details of PhiX and cluster density, etc recorded.
The first steps of processing were all done using the BioHPC cloud computers. I’m reproducing the list of commands that I entered here with some descriptions, though actually running these commands would require putting all the files up to the cloud and navigating to run each command from the right directory.
Trim Galore!using this command:
trim_galore *_R1.fastq.gz --rrbs --fastqc -o output/
I looked at a subset of the fastqc files produced by
Trim Galore! and they all looked very good in terms of quality. I’m not sure if there is something from that output we should report quantiatively, or if it’s just enough to say ‘they looked good’.
Bismarkv0.16.1 for bisulfite alignment. There are some details on setting up the BioHPC environment to have bismark in the path that I won’t list here. The first real command was to index the tree swallow genome that Leo had assembled.
bis.align.sh. The command for a single sample was:
bismark --multicore 4 /directory/bismark_genome file_trimmed.fq.gz
bis_extract.shand the command for a single file was:
bismark_methylation_extractor --single-end --multicore 4 --gzip --bedGraph file_bt2.bam
bismark2summary. The output from #4 above was used as the input for
bis_meth_convert.shand are identical to the kit directions.
Figure x. Estimates of methylation conversion from bisulfite treatment for unmethylated (A) and methylated (B) spike-in controls. Histograms show values for each sample, blue lines indicate the average across all samples, and red dashed lines show the kit suggested average targets to indicate that conversion was effective.
NOTE: I don’t think this figure needs to be included anywhere. We can just report the average value (blue line) for the methylated and unmethylated controls and say that they are within the zone recommended by the kit. I also am not sure it is really anything to worry about that a few samples are outside those zones. I read a thread by the author of Bismark suggesting that real samples typically convert better than these spike in controls and that as long as it is reasonably close to expectation there isn’t any major reason to worry about it.