Drug - Cytokine Combinatorial Screen

Welcome to the Shiny App to Explore the Drug-Cytokine Combinatorial Screen Dataset

Please click on the tabs on the left to interactively explore the data. This app accompanies the 2021 paper: Drug-microenvironment perturbations reveal resistance mechanisms and prognostic subgroups in CLL. If you use this app to support published research, please cite the paper. The data can also be explored manually using the online vignette.

Drug and stimulus responses

Explore log-transformed viabilites for single and combinatorial treatments and annnotate samples by genetic features

Effects of mutations on drug & stimulus responses

View log-transformed viability data stratified by genetic features

Genetic predictors of drug & stimulus responses

View predictor profiles for viabilities after treatment with drugs and stimuli

Genetic predictors of drug & stimulus interactions

View predictor profiles for the size of interactions between drugs and stimuli

Maintenance

This shiny app is maintained by Holly Giles. If you have questions, please email me at holly.giles@embl.de.

Security and Licence

Select a drug : stimulus combination to view viabilities for all patient samples, coloured by chosen genetic feature.

Drug:

Stimulus:

Genomic Feature:

Viabilities with single and combinatorial treatments:

For mutated samples only:

For unmutated samples only:

Beeswarm plot of log transformed viability values, stratified by chosen feature(s)

Select inputs Select the drug and/or stimulus to visualise log transformed viability values stratified by the selected genetic feature. Optional: Select a second genetic feature to visualise both simultaneously.

Drug:

Stimulus:

Genetic Feature 1:

Genetic Feature 2 (optional):

Select drug and/or stimulus to view genetic predictors of viability with selected treatment.

Drug:

Stimulus:

Adjust Frequency Threshold (the proportion of cases in which a predictor must be significant with each bootstapped repeat of the regression) and Coefficient Threshold (minimum value of predictors).

Frequency Threshold:

Coefficient Threshold:

Lasso Regularised Regression to identify genetic features that predict viability with selected treatment

The predictor profile plot depicts genetic features that predict response to selected treatment. Bar plot on left indicates size and sign of coefficients for the named predictors. Positive coefficients indicate higher viability after stimulation, if the feature is present. Scatter plot and heatmap indicate how each genetic feature relates to patient sample viabilities: Scatter plot indicates log(viability) values, in order of magnitude, for each individual sample. Heatmap shows patient mutation status for each of genetic predictors for corresponding sample in scatter plot.

Select drug and stimulus to view genetic predictors for the size of interaction (beta int) between chosen combination.

Drug:

Stimulus:

Adjust Frequency Threshold (the proportion of cases in which a predictor must be significant with each bootstrapped repeat of the regression) and Coefficient Threshold (minimum value of coefficients).

Frequency Threshold:

Coefficient Threshold:

Lasso Regularised Regression to identify genetic features that predict the size of drug - stimulus interactions.

Predictor profile plot depicts genetic features that modulate the size of interaction between chosen drug and stimulus. To generate predictor profile, a linear model was fitted in a sample - specific manner, to calculate drug- stimulus interaction coefficients (beta int) for each patient sample. Ranked patient-specific beta int values are shown in lower scatter plot. Associations between the size of beta int and genetic features were identified using multivariate regression with L1 (lasso) regularisation, with gene mutations and IGHV status as predictors. The horizontal bars on left show the size of fitted coefficients assigned to genetic features, showing those that meet selected cut offs. Matrix above scatter plot indicates patient mutation status for the selected genetic features. Matrix fields correspond to points in scatter plot (ie patient data is aligned), to indicate how the size of beta int varies with selected genetic feature.