In shbrief/PCAGenomicSignatures: Interpretation of RNA-seq experiments through robust, efficient comparison to public databases

knitr::opts_chunk$set(echo = FALSE)

RAVs best represents your dataset

The validation provides a quantitative representation of the relevance between your dataset and RAVs. Below shows the top 6 validated RAVs and the complete result is saved as {input_name}_validate.csv.

head(params$val_all)

Heatmap Table

heatmapTable takes validation results as its input and displays them into a two panel table: the top panel shows the average silhouette width (avg.sw) and the bottom panel displays the validation score.

heatmapTable can display different subsets of the validation output. For example, if you specify scoreCutoff, any validation result above that score will be shown. If you specify the number (n) of top validation results through num.out, the output will be a n-columned heatmap table. You can also use the average silhouette width (swCutoff), the size of cluster (clsizecutoff), one of the top 8 PCs from the dataset (whichPC).

Here, we print out top r params$numOut validated RAVs with average silhouette width above 0.

heatmapTable(params$val_all, num.out = params$numOut, swCutoff = 0)

Sample Score

RAV-assigned scores for each sample can be used to compare the features represented by the given RAV across different datasets. Below shows a part of sample scores in a heatmap, where scores are assigned to each sample (row) from each RAV (column). The complete result is saved as {input_name}__sampleScore.csv.

k <- min(nrow(params$score), 10)
sampleScoreHeatmap(params$score[seq_len(k),], 
                   dataName = params$inputName, 
                   modelName = "RAV",
                   column_names_gp = 10)

Interactive Graph

Under the default condition, plotValidate plots validation results of all non single-element RAVs in one graph, where x-axis represents average silhouette width of the RAVs (a quality control measure of RAVs) and y-axis represents validation score. We recommend users to focus on RAVs with higher validation score and use average silhouette width as a secondary criteria.

plotValidate(params$val_all, interactive = TRUE)

Note that interactive = TRUE will result in a zoomable, interactive plot that included tooltips, which is saved as {input_name}_validate_plot.html file.

You can hover each data point for more information:

• sw : the average silhouette width of the cluster
• score : the top validation score between 8 PCs of the dataset and RAVs
• cl_size : the size of RAVs, represented by the dot size
• cl_num : the RAV number. You need this index to find more information about the RAV.
• PC : test dataset's PC number that validates the given RAV. Because we used top 8 PCs of the test dataset for validation, there are 8 categories.

If you double-click the PC legend on the right, you will enter an individual display mode where you can add an additional group of data point by single-click.

Prior information associated to your dataset

validated_ind <- validatedSignatures(params$val_all, num.out = params$numOut,
                                     swCutoff = 0, indexOnly = TRUE)

# In case, there are fewer validated_ind than the number of outputs user set
n <- min(params$numOut, length(validated_ind), na.rm = TRUE)

MeSH terms in wordcloud

for (i in seq_len(n)) {
    set.seed(1)
    print(paste0("MeSH terms related to RAV", validated_ind[i]))
    drawWordcloud(params$RAVmodel, validated_ind[i])
}

GSEA

The complete result is saved as {input_name}_genesets_RAV*.csv.

res_all <- vector(mode = "list", length = n)
for (i in seq_len(n)) {
    RAVnum <- validated_ind[i]
    RAVname <- paste0("RAV", RAVnum)
    res <- gsea(params$RAVmodel)[[RAVname]]
    res_all[[i]] <- head(res)
    names(res_all)[i] <- paste0("Enriched gene sets for RAV", validated_ind[i])
}
res_all

Publication

The complete result is saved as {input_name}_literatures_RAV*.csv.

res_all <- vector(mode = "list", length = n)
for (i in seq_len(n)) {
    RAVnum <- validated_ind[i]
    res <- findStudiesInCluster(params$RAVmodel, RAVnum, studyTitle = TRUE)
    res_all[[i]] <- head(res)
    names(res_all)[i] <- paste0("Studies related to RAV", validated_ind[i])
}
res_all

shbrief/PCAGenomicSignatures documentation built on May 3, 2023, 11:21 a.m.