The application for predicting potential effective combination treatments for melanoma patients. Here is a brief description for each tab.

(1) Top combinations
(2) PC1 and survival
(3) TME (Tumor immune microenvironment)
(4) Mutation & survival
(5) eQTL.

The top combinations tab contains two tables of the predicted top combinations. Oned based on expression results ranking and one on mutation results.
The prediction method used for ranking the combination is described in the manual and manuscript.

The PC1 & survival tab shows the tables of genes and PC1 loadings in the target gene sets of each treatment
for each cohort along with the KM plots of PC1 and overall survival for each treatment in both TCGA and Moffitt cohorts.
The PCs are dichotomized at the median.

The TME tab contains tumor microenvironment heatmaps (both for all target genes and a single gene) and violin plots that display
the summarized percent cells with expression for genes within a drug target gene set for each treatment in different cell types
from a melanoma single cell dataset. Single gene option is also available. The single cell data for the heatmaps and violin plots
come from forty-one human melanoma specimens and 2 normal cerebrospinal fluid samples from 2 non-leptomeningeal
melanoma metastases patients. More information can be found in our previous publication.

The mutation & survival tab displays Kaplan-Meier plots for the association between overall survival and mutation status
in the target genes of the two treatments and the combination of the mutation status of each
treatment in each cohort.

The eQTL tab displays the box plots of gene expression in both cohorts by mutation status in the target genes.
It also displays the summary statistics of the debatched expression on log scale.

Contact information: ann.chen@moffitt.org, zachary.thompson@moffitt.org

Top combinations: expression based

Loading...

Loading...
Loading...

Loading...
Loading...
Loading...
Loading...
Loading...
Loading...

Loading...
Loading...

Loading...
Loading...
Loading...
Loading...

Loading...
Loading...
Loading...
Loading...
Loading...
Loading...

Loading...
Loading...

Loading...
Loading...
Loading...

Top combination ranking method

The method for ranking combinations uses Cox PH regression models for both cohorts with the first principal
component of the expression of the target genes of each drug. Fishers product method was employed to
combine the p values of the 4 main effects (two from each cohort) when fitting the Cox expression
regression models from interaction models, to determine the top combinations we filtered on
p values (< 0.05) of the single drug models.

A note about PCA results

If the signs of the loadings of the principle components between Moffitt and TCGA disagree in more than 50% of them
the orientation of the PC1 in TCGA is flipped. The PC1 of one or both drugs may be flipped

PD1-PDL1 - Target genes and loadings in TCGA - original orientation

PD1-PDL1 - Target genes and loadings in TCGA - PC1 orientation flipped

Loading...
Loading...
Loading...
Loading...