Chr instability: PR 3 of 3: Histology plots #532
Conversation
…dea why cnv heatmap wasn't being updated????
|
Alrighty, @jashapiro , let me know if we think everything is addressed now. |
analyses/chromosomal-instability/02b-plot-chr-instability-by-histology.Rmd
Show resolved
Hide resolved
analyses/chromosomal-instability/02b-plot-chr-instability-by-histology.Rmd
Show resolved
Hide resolved
| @@ -356,7 +358,7 @@ breaks_density_list <- lapply(breaks_list, function(breaks_df) { | |||
| samples, experimental_strategy, genome_size | |||
| ) %>% | |||
| # Count number of mutations for that sample | |||
| dplyr::summarize(breaks_count = dplyr::n()) %>% | |||
| dplyr::summarize(breaks_count = sum(!is.na(chrom))) %>% | |||
There was a problem hiding this comment.
I just realized, while thinking about #490 (comment), that this may make some samples that should be NA into zeros. If the sample was not in the consensus seg file, it should be NA for CNV breaks and SV breaks I think such a sample here would end up with a zero here.
There was a problem hiding this comment.
Those do end up with zeroes. If NA is preferred I can make those changes. I’ll just have to convert them back to zeroes for the CDF plots, unless we think those samples should be dropped from the CDF plots?
There was a problem hiding this comment.
Yes, drop them if they are missing. If you don't drop them, it says to your audience that there are n samples that have 0 breakpoints when we don't have evidence either way.
There was a problem hiding this comment.
So is it correct to say that all WGS samples have been ran through both CNV detection pipelines, so if they don't show up in consensus CNV file or the SV file then they should be NAs for break density?
There was a problem hiding this comment.
If I'm understanding this correctly, then there will be no such thing as a 0, only NAs and the minimum break_counts would 1.
There was a problem hiding this comment.
@cansavvy take a look at the documentation for https://github.com/AlexsLemonade/OpenPBTA-analysis/tree/master/analyses/copy_number_consensus_call and also see https://github.com/AlexsLemonade/OpenPBTA-analysis/blob/master/analyses/copy_number_consensus_call/results/uncalled_samples.tsv. I would also check the IDs in the original files to be sure.
There was a problem hiding this comment.
The docs linked to above by @jaclyn-taroni should explain, but yes, there are both zeros and NAs. Some samples were deemed “uncallable” and should be NA. Others were called but had no CNVs and should be 0.
| dplyr::summarize(is_na = any(is.na(chrom)), | ||
| breaks_count = dplyr::n()) %>% | ||
| # Calculate breaks density, but put NA for breaks_count if the sample was not | ||
| # in the SV or CNV data originally | ||
| dplyr::mutate(breaks_count = dplyr::case_when( | ||
| !is_na ~ as.numeric(breaks_count), | ||
| is_na ~ as.numeric(NA) | ||
| ), |
There was a problem hiding this comment.
Unfortunately, this isn't actually doing what we really want it to do.
The trouble is that anything with a true zero count will get an NA in chrom column, just like something that is actually missing.
I think the thing to do is to add a column to metadata called surveyed which would indicate if the sample is in unique(c(cnv_samples, sv_samples)). Then if surveyed is TRUE, set the is_na break_counts to 0, and to NA otherwise.
something like:
dplyr::mutate(
breaks_count = dplyr::case_when(
!is_na ~ as.numeric(breaks_count),
is_na & surveyed ~ as.numeric(0),
TRUE ~ as.numeric(NA)
), |
Okay, I think this is good to go. |
AlexsLemonade#532 changed results.
* Initial files added to ependymoma subtyping folder * Added bash script and changed the paths for all files to run from OpenPBTA directory * Added bash script and changed the paths for all files to run from OpenPBTA directory * Add Ependymoma subtyping to CI * Add to analyses/README.md * Test removing rpy2 and using pyreadr exclusively * Change to use pyreadr properly * Revert pyreadr changes * Add subset flag to CI * Use R to generate subset file & shell to specify filenames * Add results file * Move Ependymoma subtyping up in CI * Responding to pull request reviews * Adding jupyter notebook * Typo fixes * Update gistic filename * Changed implemented as suggested on feb 7 2020 * Small review changes remove unused imports simplify use of `stats` * Update 00-subset-for-EPN.R with changes from @cansavvy code review * Zscore column names changed * Changed how merge is done between RNA and DNA tables * Removed comment lines * remove duplicate commented code. * Added some columns as per comments from 02-12-2020 * update invocation of 02_ependymoma_generate_all_data.py Add line continuation characters to bash script Remove no longer used --breakpoints option * Handle missing data, and some refactoring I made some substantial changes here in structure, but the results should be largely unchanged. I did some transposing when constructing data frames so we can use the same function (fill_df) to extract data more often, and moved the ID column specification out of the function so that RNA and DNA-derived data are handled the same way. The function then allows a set of samples to be specified, and if the request is for a a sample that does not fall in there, it is set as NA for that column in the output data, otherwise it is filled in with a default value. * Delete unused full table zscore * Rerun with updated data #532 changed results. Co-authored-by: jashapiro <jashapiro@gmail.com>
Purpose/implementation Section
This includes redone versions of the material that was included in #492.
This last PR has the last of three notebooks which contain the material that was originally all in the previous 01 notebook.
What scientific question is your analysis addressing?
How does chromosomal instability relate to histology group?
What was your approach?
This takes the binned counts from the 01-localization notebook and makes histology plots for them.
What GitHub issue does your pull request address?
#487
Directions for reviewers. Tell potential reviewers what kind of feedback you are soliciting.
Which areas should receive a particularly close look?
Most of this material has been reviewed previously, but it now is in its own notebook.
How do you feel about the readability of it? Are there other aesthetic changes that need to be made to the plots?
What is your summary of the results?
Here's the rendered html: https://cansavvy.github.io/openpbta-notebook-concept/chromosomal-instability/02b-plot-chr-instability-by-histology.nb.html
Reproducibility Checklist
These items were done previously.
Documentation Checklist
These items were done previously.
READMEand it is up to date (Note it includes documentation for the upcoming notebooks as well.analyses/README.mdand the entry is up to date.