Run mixomics and two simple files

[sr320]

Using gene expression https://github.com/epigeneticstoocean/2018_L18-adult-methylation/blob/main/analyses/gene_fpkm.csv

gene level methylation https://raw.githubusercontent.com/sr320/ceabigr/main/output/40-gene-methylaiton.csv

[kubu4]

@sr320 - Do you know why the gene level methylation file has NAs present?

[kubu4]

To add some background, I was messing around with the Mixomics package trying to run a PCA with the gene methylation file and the PCA function would just run indefinitely (whereas the gene expression file would load almost instantly). I was able to figure out that the difference between the two files was NAs in the gene methylation file. After replacing those NAs with 0, the PCA function worked.

[sr320]

NAs are likely the result of having genes that have CpGs with no methylation data. Or for that matter it could be a result of taking a mean of all CpGs in a gene where at least one was a NA (not sure how NAs would be handled in that instance.)

[kubu4]

Thanks. Interesting.

So, if we have no methylation data for a gene, does that mean an average methylation of 0 or does that mean had insufficient sequencing depth across the gene to reach the set threshold (e.g. < 5x coverage across entire gene)?

[sr320]

Insufficient sequencing depth

On Wed, Aug 31, 2022 at 3:59 PM kubu4 @.***> wrote:

Thanks. Interesting.

So, if we have no methylation data for a gene, does that mean an average methylation of 0 or does that mean had insufficient sequencing depth across the gene to reach the set threshold (e.g. < 5x coverage across entire gene)?

— Reply to this email directly, view it on GitHub [github.com] https://urldefense.com/v3/__https://github.com/sr320/ceabigr/issues/68*issuecomment-1233362726__;Iw!!K-Hz7m0Vt54!jO3JTe8j8gJDyy_xjQAj_sqsFdG1ykp5YRLPoaFURO_-mEsuJbT9LMiAMBslNINLOAOpuEzG-UVu6aAh8a2sTg4$, or unsubscribe [github.com] https://urldefense.com/v3/__https://github.com/notifications/unsubscribe-auth/ABB4PNY2JGKSAJPDPGE34WLV362STANCNFSM6AAAAAAQAS5ABY__;!!K-Hz7m0Vt54!jO3JTe8j8gJDyy_xjQAj_sqsFdG1ykp5YRLPoaFURO_-mEsuJbT9LMiAMBslNINLOAOpuEzG-UVu6aAh-8SI6ts$ . You are receiving this because you were mentioned.Message ID: @.***>

--

Steven B. Roberts, Associate Professor Associate Director - Graduate Program Coordinator School of Aquatic and Fishery Sciences University of Washington Fisheries Teaching and Research (FTR) Building - Office 232 1140 NE Boat Street - Seattle, WA 98105 robertslab.info - @.*** - @sr320 vm:206.866.5141 - cell:360.362.3626 schedule a zoom call: https://d.pr/gsgxVJ

[kubu4]

I decided to mess around with this a bit, but not extensively so I don't have a dedicated project for this. Here's rendered R Markdown if anyone's interested:

https://rpubs.com/kubu4/cvir-gonad-oa-mixomics_testing

Basically, ran the following:

Average gene methylation

• PCA
• sparse PCA (sPCA)
• identified top ~25 genes responsible for defining each principal component

Both PCA and sPCA show clustering by sex, regardless of methylation status and/or treatment.

Average gene methylation and gene expression (FPKM)

• PLS (Projection to Latent Structure) and sparse PLS (sPLS)
• show tight clustering by sex, with no noticeable impact from treatment
• Identifies to ~20 genes contributing to the two sPLS components.