dgrun / fateid

R package for the inference of cell fate bias from single-cell RNA-seq data.

FateID algorithm

FateID is a random forests-based algorithm for the quantification of the fate biases of multipotent progenitors, i. e. the probabilities of differentiating towards particular cell lineages. It utilizes single cell expression data, and was designed to work well with quantitative single-cell RNA-seq data incorporating unique molecular identifiers. FateID requires a gene-by-cell expression matrix as input and groups of cells represeting committed stages (or advanced differentiation stages) of all lineages accessible from the multipotent progenitor population. These so-called target clusters can be defined by a clustering partition, which can also directly be generated by FateID if a list of specific marker genes for all lineages is given.

Installing

After downloading and unzipping

unzip FateID-master.zip 

it can be installed from the command line by

R CMD INSTALL FateID-master

or directly in R from source by

install.packages("FateID-master",repos = NULL, type="source")

(if R is started from the directory where FateID.zip has been downloaded to; otherwise specify the full path)

Alternatively, install in R directly from github using devtools:

install.packages("devtools")
library(devtools)
install_github("dgrun/FateID")

Running a FateID analysis

Load package:

library(FateID)

See vignette for details and examples:

vignette("FateID")

Reference:

Herman JS, Sagar, Grün D. (2018) FateID infers cell fate bias in multipotent progenitors from single-cell RNA-seq data. Nat Methods. 2018 May;15(5):379-386. doi: 10.1038/nmeth.4662.

fateid's Issues

Importing from Seurat

Dear Developers,
I noticed in the paper (Herman, Sagar, and Grün 2018) that Seurat was used also. Have you ever tried to import a seurat object into fateID? If you have, i could work in this way and adjust the script accordingly by giving consensus genes of interest. I would like to know how I can do this (if its not alot of work) and if there was ever a script around to do this.
Thanks a lot.

princurve 2.0.2 has been released

Hello Dominic @dgrun!

I just updated princurve to version 2.0.2 on CRAN. I noticed the FateID package uses princurve and wanted to notify you of the changes made so you can update your package accordingly.

The principal.curve function has been renamed to principal_curve, in order to get princurve up-to-date with current R naming conventions. In order to make the names of the parameters more self explanatory, the argument plot.true has been renamed to plot_iterations, and the output variable tag has been renamed to ord.

The function principal.curve still exists in the package, for now, as not to break your package. On July 1st, however, this function will become deprecated, and will later on be removed.

If you have any remarks or questions, feel free to let me know!
Robrecht

The average pseudo-temporal expression profile of genes within individual modules of the SOM

Hi, thank you for developing FateID!
I ran the code as the reference manual provided with default parameters. But when I plot the average pseudo-temporal expression profile of genes within individual modules of the SOM, I noticed some variances in my plot. How can I optimize my parameters?
Thank you so much!!!

Running FateID on seurat object issues

Hi,

Thank you for developing FateID. I have used the vignette on the original data with no problem. Now, I would like to use FateID in my data set. I have processed my data with Seurat and have a final object on which I want to identify fate biases. I have the following questions;

1. Is there now a function available to import directly from Seurat?
2. If I import my data from Seurat manually, am I right to say that "x" is the un-normalized counts matrix; and "y" will be cell identities? What about "v"?
3. If I reclassify all remaining cells using the cells within the target clusters as input with FateID, then I do not have to reclassify cells using the full data frame containing all genes can be used as input? (i.e. steps below)

4. If I reclassify all remaining cells using the cells within the target clusters as input with FateID, do I still need to do feature selection using the _getFeat_ function? 5. I ran into an error while computing the fate bias using _fateBias_ function. The error is the following:

Error in randomForest.default(xr, as.factor(pr), xt, nbtree = nbtree, :
Need at least two classes to do classification.

After this, I am stuck and cannot proceed. Any help will be much appreciated.

Thanks!