R - Need a virtual session on list manipulation

[kubu4]

I'm mired in R list junk and am in desperate need of someone to walk through how to deal with/understand lists. I find it confusing on how to deal with lists containing different type of objects (e.g. a list of data frames) and how to work with them.

The end goal is extract/combine data and determine row-wise coefficients of variation.

Are there any list experts who'd be willing/interested in spending some time walking through my project and providing me with some guidance/explanation/help?

Science hour would be ideal. Otherwise, maybe we can explore some of this at the next lab meeting. Or, "worst case" I'm also up for a dedicated session specifically for this at some other time.

In the mean time, I'll continue to do some reading/practicing on the topic and see if I can get a better grasp of things.

The other benefit of the session might also reveal some different approaches that other people would take which would be significantly less confusing!!!

For kicks, I'll post some of the stuff I'm working on in this discussion, in case anyone wants to peruse it. However, it quickly gets a bit complicated to explain without walking through the entire process, thus my interest in discussing via video chat.

[kubu4]

Git Repo: https://github.com/epigeneticstoocean/2018_L18-adult-methylation

R Markdown file: https://github.com/epigeneticstoocean/2018_L18-adult-methylation/blob/main/code/ballgown_analysis.Rmd

I'm working with the gene expression analysis package ballgown. It creates a custom object that is, essentially, a large database of all the expression values for all of your samples, at the transcript level. It also includes info on exons/introns/genes, as well as metadata the user provides it with (e.g. which samples are female/male, which are control/exposed, etc).

---

Part of code that I'm trying to wrangle begins here:

https://github.com/epigeneticstoocean/2018_L18-adult-methylation/blob/667495d5551c3e47a844724fc74372ad171eb5eb/code/ballgown_analysis.Rmd#L720

---

ballgown then makes it very easy to extract just a subset of the samples (in this case, from the TreatmentN metadata):

https://github.com/epigeneticstoocean/2018_L18-adult-methylation/blob/667495d5551c3e47a844724fc74372ad171eb5eb/code/ballgown_analysis.Rmd#L739

This is part where it is difficult to get the information into a text format without perusing the data with someone else present. Since it's an object, there's a lot of information associated with it that I can't easily copy/paste here.

---

Next, I can easily extract transcript expression (FPKM) data from that subset of samples (for looping purposes, I pop this info into a list):

https://github.com/epigeneticstoocean/2018_L18-adult-methylation/blob/667495d5551c3e47a844724fc74372ad171eb5eb/code/ballgown_analysis.Rmd#L742

Here's what one entry in the list looks like:

treatment_fpkm_list[[1]]

        FPKM.S13M    FPKM.S64M    FPKM.S6M    FPKM.S7M
1       63.109000  1161.543557  110.668782  239.215973
2      158.217649 82874.949689  487.061698 2602.800279
3       72.036519   963.764924   93.048271  241.462390
4       55.182350   107.488847   52.721938  127.347810
5       97.711458   154.764867   84.386166  145.411473

---

I also have a data frame where the t_id column matches the row numbers shown in the table above:

I'd like to "merge" this data frame and the info from the list above, keeping only the columns in that list and adding gene_name from the data frame.

Or, a different approach would be to use the contents of the list to select the desired columns from the data frame. Actually, this might be a better approach?

Any suggestions for this step?

[yaaminiv]

I'm 90% sure I've dealt with something similar before........let me do some digging

[yaaminiv]

@kubu4 I can't find my code that I thought I had somewhere, but my first instinct is to save the list as a data frame, then merge. It seems like you're already doing that?

https://github.com/epigeneticstoocean/2018_L18-adult-methylation/blob/667495d5551c3e47a844724fc74372ad171eb5eb/code/ballgown_analysis.Rmd#L760

[kubu4]

I'm getting closer. Need/want to add column to items in list from a vector of gene names. Here's what I did:

# Add column called "gene_name", comprised of gene names, to existing list item
> treatment_fpkm_list[[1]] <- cbind(treatment_fpkm_list[[1]], gene_name = whole_tx_table_gene.names)
> treatment_fpkm_list[[1]]

This adds a column and makes the list item look like this:

      FPKM.S13M     FPKM.S64M      FPKM.S6M      FPKM.S7M      gene_name                     
1     "63.109"      "1161.543557"  "110.668782"  "239.215973"  "COX1"                        
2     "158.217649"  "82874.949689" "487.061698"  "2602.800279" "rna-NC_007175.2:1710..8997"  
3     "72.036519"   "963.764924"   "93.048271"   "241.46239"   "COX3"                        
4     "55.18235"    "107.488847"   "52.721938"   "127.34781"   "rna-NC_007175.2:3430..3495"  
5     "97.711458"   "154.764867"   "84.386166"   "145.411473"  "rna-NC_007175.2:3499..3567"  
6     "82.522413"   "154.972338"   "67.927814"   "120.346216"  "rna-NC_007175.2:3578..3646"

[laurahspencer]

Have you tried the ‘bind_rows’ function? For example I used it in the below code to combine data frames saved to the list object “DML.features”, except for the 7th df in the list.

DML.features.df <- bind_rows(DML.features[-7])

[kubu4]

Have you tried the ‘bind_rows’ function?

I have not. Thanks, this is intriguing...

[kubu4]

I have figured out some stuff in ballgown that may make this Discussion entirely moot. Will report back with more deets in a bit...

[kubu4]

Thanks to those that chimed in. Even though I made this a Discussion about list manipulation, it was really a specific issue in manipulating ballgown objects and tying into using the rowcvs() function.

For posterity, I'll post the true underlying issue that was screwing me up.

Let's say one has created a ballgown object, called bg. To extract transcript info, one would run this command:

texpr(bg)

If one wants to extract gene info, one would run this command:

gexpr(bg)

Although these commands, have virtually identical syntax, there is a critical difference in what is output.

texpr(bg) spits out a ballgown object.

gexpr(bg) spits out a matrix.

I was trying to manipulate the outputs from both commands, expecting a ballgown object from each. Since I was unable to, that lead me to trying to work solely with the object output from texpr(bg) and trying to perform manipulations of data at the gene level. This greatly overcomplicated things.

Once I figured out that gexpr(bg) output a matrix, I was able to easily incorporate this into my existing code.