urszulaczerwinska / deconica Goto Github PK
View Code? Open in Web Editor NEWDeconvolution of transcriptome through Immune Component Analysis
Home Page: https://urszulaczerwinska.github.io/DeconICA/
License: GNU General Public License v2.0
Deconvolution of transcriptome through Immune Component Analysis
Home Page: https://urszulaczerwinska.github.io/DeconICA/
License: GNU General Public License v2.0
The idea is to add fastGSEA as an alternative to fisher test
implement t.test on A matrix selected samples
save sample names, it will be important for t-test
Instead of using enrichment-based test we can use functional interpretation to interpret components
missing unit tests :
missing examples :
Sometimes we get different number of immune component candidates (stroma ones don’t always pass the threshold) - possibility: not take them into account for deconvolution
R check is failing on unit test that works fine from the unit module, it is not seeing data in data/.rda
write clean function and place in pipeline for FEV
correlation with LM22 indicates other candidates
Some errors have slipped into the tutorial, making some commands not usable with a simple Copy-Pasting.
Here is a list of some of which I have found, assigned by categories:
Issue 1:
No folder "data-raw" can be found in deconICA package from R. It is present in the GitHub repository but during the installation in R, the folder is not present.
Some commands to not work, such as:
GE_SDY420 <- read.delim("./data-raw/xCell_ImmPort/GE_SDY420.txt", row.names=1, stringsAsFactors=FALSE)
FACS_SDY420 <- read.delim("../data-raw/xCell_ImmPort/FCS_SDY420.txt", row.names=1, stringsAsFactors=FALSE)
TIMER <- ACSNMineR::format_from_gmt("./data-raw/TIMER_cellTypes.gmt")
data(BEK_ica_overdecompose)
Issue 2:
When assigning metagenes, the whole matrix should be given as input without the "$r".
This is only found in the html version of the tutorial (not the Rmd one).
Here are the cases where this happens:
mix1.assign.20 <- assign_metagenes(mix1_corr.basis.20$r, exclude_name = NULL)
GE_SDY420_ica_39.LM22.reciprocal.corr <- assign_metagenes(GE_SDY420_ica_39.corr.LM22$r, exclude_name = NULL)
reciprocal.corr <- assign_metagenes(corr$r, exclude_name = NULL)
reciprocal.corr.Biton <-assign_metagenes(corr_Biton$r, exclude_name = c("M8_IMMUNE", "M2_GC_CONTENT"))
Issue 3:
When using "kable", no import has been stated prior to the usage, creating an error.
library(kableExtra) should be mentioned at some point in the tutorial.
Issue 4:
When generating html tables with "kable", the command to visualise the table in Rstudio should be added as some cases lack it. For instance, the line
kable(GE_SDY420_ica_39.LM22.reciprocal.corr, "html", row.names = FALSE)
should become
kable(GE_SDY420_ica_39.LM22.reciprocal.corr, "html", row.names = FALSE) %>% kable_styling(font_size = 8)
Issue 5:
Some typos here and there are hidden. Here would be the needed correction imo:
cell_prop <- pData(GSE64385)[ , c(1, 2, 10, 11,12, 13, 14, 15, 16, 17)]
overdecompose parameter that selects number of composed components? needed to perform overdecomposition of the input matrix
An efficient way to in interpret a component is to use correlation with some known profile or as we call it a
type (a metagene), we can then correlate them with obtained components and verify if some of decomposed
They can be considered as a negative control.
#without endothelial cells and Fibroblasts
It will be necessary in the final version of the plugin to have the lolypop/radar/rcorr plot the labels for identified components
It would be necessary to introduce a way to make the enrichment unsensible to the gene threshold
by scanning many thresholds and selecting the most voted result among all thresholds
I repeated the pipeline on TCGA I downloaded and transformed counts in log2 and results are really not interpretable by enrichment
It would be good to find faster way for enrichment (figure out what is long in Paul's function and rewrite Paul's function)
test on Windows and on Unix
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.