jinmiaochen
commented
7 years ago Hi Erin,
Thanks for your feedback. Our intension is to use selected parameters for clustering, but allow users to view the expression of all markers. We will look into it and get back to you.
Best, Jinmiao
From: Erin Simonds [mailto:notifications@github.com] Sent: Wednesday, October 11, 2017 12:56 PM To: JinmiaoChenLab/cytofkit cytofkit@noreply.github.com Cc: Subscribed subscribed@noreply.github.com Subject: [JinmiaoChenLab/cytofkit] All parameters are used for PhenoGraph, yielding useless clusters (#11)
The current version of Cytofkit on Bioconductor (1.8.3) appears to use all parameters for the PhenoGraph clustering step, causing very uninformative clusters (see attached PDF). Also, it throws an error when performing PhenoGraph clustering, whereas previous versions (as late as 1.8.1) did not.
Screenshots showing PhenoGraph bug in Cytofkit v183.pdfhttps://github.com/JinmiaoChenLab/cytofkit/files/1374239/Screenshots.showing.PhenoGraph.bug.in.Cytofkit.v183.pdf
Testing:
I used the FCS file included with the R package ("130515_C2_stim_CD19-.fcs") for testing, and I compared v1.8.1 to v1.8.3 with the same analysis parameters.
Console output from 1.8.3:
`
library(cytofkit) Loading required package: ggplot2 Loading required package: plyr cytofkit_GUI() Input arguments:
- Project Name: cytofkit
- Input FCS files for analysis: -130515_C2_stim_CD19-.fcs
- Markers: -(Cd112)Di -(La139)Di -(Nd142)Di -(Nd146)Di -(Sm154)Di -(Gd156)Di
- Data merging method: ceil
- Data transformation method: cytofAsinh
- Dimensionality reduction method: tsne
- Data clustering method(s): Rphenograph
- Data visualization method(s): tsne
- Subset progression analysis method: NULL
Extract expression data... 5000 x 52 data was extracted! Dimension reduction... Running t-SNE...with seed 42 DONE Run clustering... Running PhenoGraph...Run Rphenograph starts: -Input data of 5000 rows and 52 columns -k is set to 30 Finding nearest neighbors...DONE ~ 2.628 s Compute jaccard coefficient between nearest-neighbor sets...DONE ~ 2.102 s Build undirected graph from the weighted links...DONE ~ 0.424 s Run louvain clustering on the graph ...DONE ~ 0.362 s Run Rphenograph DONE, took a total of 5.51599999999996s. Return a community class -Modularity value: 0.8562154 -Number of clusters: 18 DONE! Progression analysis... Listing markers used for dimension reduction... Wrapping results... Analysis DONE, saving the results... R object is saved in cytofkit.RData THIS R OBJECT IS THE INPUT OF SHINY APP! Save to file: /Users/esimonds/Downloads/130515_C2_stim_CD19-.fcs Writing results Done! Results are saved under path: /Users/esimonds/Downloads Warning message: In if (!(right_marker)) { : the condition has length > 1 and only the first element will be used`
Console output from 1.8.1:
`Restarting R session...
library(cytofkit) Loading required package: ggplot2 Loading required package: plyr packageVersion("cytofkit") [1] ‘1.8.1’ cytofkit_GUI() Input arguments:
- Project Name: cytofkit_181
- Input FCS files for analysis: -130515_C2_stim_CD19-.fcs
- Makrers: -(Cd112)Di -(La139)Di -(Nd142)Di -(Nd146)Di -(Sm154)Di -(Gd156)Di
- Data merging method: ceil
- Data transformation method: cytofAsinh
- Dimensionality reduction method: tsne
- Data clustering method(s): Rphenograph
- Data visualization method(s): tsne
- Subset progression analysis method: NULL
Extract expression data... 5000 x 6 data was extracted! Dimension reduction... Runing t-SNE...with seed 42 DONE Run clustering... Runing PhenoGraph...Run Rphenograph starts: -Input data of 5000 rows and 6 columns -k is set to 30 Finding nearest neighbors...DONE ~ 0.124 s Compute jaccard coefficient between nearest-neighbor sets...DONE ~ 2.411 s Build undirected graph from the weighted links...DONE ~ 0.414 s Run louvain clustering on the graph ...DONE ~ 0.318 s Run Rphenograph DONE, totally takes 3.267s. Return a community class -Modularity value: 0.8043747 -Number of clusters: 15 DONE! Progression analysis... Analysis DONE, saving the reuslts... R obejct is saved in cytofkit_181.RData THIS R OBJECT IS THE INPUT OF SHINY APP! Save to file: /Users/esimonds/Downloads/130515_C2_stim_CD19-.fcs Writing results Done! Results are saved under path: /Users/esimonds/Downloads`
The tSNE maps look identical between the two versions, which tells me that the tSNE code is correctly using only the 6 parameters I specified.
The PhenoGraph clusters in v1.8.1 look good (they correlate with cell type). Notice that the console output says:
Runing PhenoGraph...Run Rphenograph starts: -Input data of 5000 rows and 6 columns
...that's all good.
The PhenoGraph clusters in v1.8.3 look almost random. Notice that the console output says:
Running PhenoGraph...Run Rphenograph starts: -Input data of 5000 rows and 52 columns
We don't want to run PhenoGraph on the entire matrix because the clusters will be defined by useless parameters, including Time and barcode parameters. Please fix! Thanks!
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHubhttps://github.com/JinmiaoChenLab/cytofkit/issues/11, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AIe92l_eNM_Ju4NY6QkV-NO53d9T4P6Mks5srEpNgaJpZM4P06mC.
This e-mail and any attachments are only for the use of the intended recipient and may contain material that is confidential, privileged and/or protected by the Official Secrets Act. If you are not the intended recipient, please delete it or notify the sender immediately. Please do not copy or use it for any purpose or disclose the contents to any other person.
esimonds
lconde-ucl
MattMyint
SamGG
Mkang1204
The current version of Cytofkit on Bioconductor (1.8.3) appears to use all parameters for the PhenoGraph clustering step, causing very uninformative clusters (see attached PDF). Also, it throws an error when performing PhenoGraph clustering, whereas previous versions (as late as 1.8.1) did not.
Screenshots showing PhenoGraph bug in Cytofkit v183.pdf
Testing:
I used the FCS file included with the R package ("130515_C2_stim_CD19-.fcs") for testing, and I compared v1.8.1 to v1.8.3 with the same analysis parameters.
Console output from 1.8.3:
Console output from 1.8.1:
The tSNE maps look identical between the two versions, which tells me that the tSNE code is correctly using only the 6 parameters I specified.
The PhenoGraph clusters in v1.8.1 look good (they correlate with cell type). Notice that the console output says:
...that's all good.
The PhenoGraph clusters in v1.8.3 look almost random. Notice that the console output says:
We don't want to run PhenoGraph on the entire matrix because the clusters will be defined by useless parameters, including Time and barcode parameters. Please fix! Thanks!