loosc/systemsserology

Machine learning tools for the analysis of systems serology data

systemsseRology

About

systemsseRology is a collection of functions used to analyse multivariate systems serology data. You can learn more about the workflow by looking at vignette("example_analysis").

Installation

systemsseRology requires the ropls package:

if (!requireNamespace("BiocManager", quietly = TRUE))
    install.packages("BiocManager")

BiocManager::install("ropls")

It can then be installed via devtools:

install.packages("devtools")
install_github("LoosC/systemsseRology", ref = "reboot")
library(systemsseRology)

Selected Papers

systemsseRology has been employed in a number of papers, including:

• Bartsch, Y. C. et. al. (2021). Viral Rebound Kinetics Correlate with Distinct HIV Antibody Features. Mbio, 12(2).

• Herman, J. D. et. al. (2021). Functional Antibodies in COVID-19 Convalescent Plasma. medRxiv.

• Bartsch, Y. C. et. al. (2021). Humoral signatures of protective and pathological SARS-CoV-2 infection in children. Nature medicine, 1-9.

• Gorman, M. J. et. al. (2021). Collaboration between the Fab and Fc contribute to maximal protection against SARS-CoV-2 in nonhuman primates following NVX-CoV2373 subunit vaccine with Matrix-M™ vaccination. bioRxiv.

• Krystle, K. Q. et. al. (2021). Comorbid illnesses are associated with altered adaptive immune responses to SARS-CoV-2. JCI insight.

systemsserology's Issues

Include real data set

So far only artificially generated data set is used. Better to use a real one (maybe the same as from the La Jolla workshop)

add handling of NA values

in example analysis add handling of NA values
maybe via knnImputation

first issue

for demonstration purposes

Include sparse PLS-DA

Instead of having LASSO + PLS-DA use the sparse version of PLS-DA (I guess its in the package mixOmics)

Helpful cheat sheet for developing R packages

https://rstudio.com/wp-content/uploads/2015/03/devtools-cheatsheet.pdf

Addition of features like polyfunctionality/breadth etc.

Meaning of ellipses on PLS-DA scores plot

Hi, thanks for making the systemsseRology package available!

I noticed a comment in the code for systemsseRology::visualize_ropls_scores() which very briefly says

# confidence levels for ellipses

The ellipse drawing function that is called is ggplot2::stat_ellipse() (documentation here).

I think the comment could be a little misleading because the ellipses being drawn are data ellipses, not confidence ellipses, so you may want to change the text to something like:

# The level at which to draw the ellipse for the data ellipse

I had a brief exchange with your colleague, Deniz Cizmeci, to make sure this detail is correct.
Thanks for your attention to this minor point. For one of our recent studies, this was a question that came up during manuscript preparation, so I wanted to bring it to your attention.

Error: two LVs required

plt_scores_pca <- visualize_ropls_scores(model_pca, y, options = opts_plot)

returns this error:

Error in visualize_ropls_scores(model_pca, sg, options = opts_plot) :
two LVs required

example univariate plots

adapt in example analysis (comparison between groups)

feature selection more than 2 classes

handle output for family "multinomial" for cv.glmnet in case of more than 2 classes in select_lasso.R

output feature_count in select_repeat.R

validate -> validate_internal

It seems a little ridiculous to have
validate(X, y, method, options)
and
validate_repeat(X, y, method, options, n_trials = 100)

We should consider renaming validate_repeat -> validate and validate -> validate_internal (not exported to users) throughout all of the code. Then users can just specify n_trials in validate(). Maybe this defaults to 1 then instead of 100?