When using prepareSummarizedExperiment there is a warning informing of the number of genes missing in the input that will be set to 0 for the model. However computeGeneSetScores always produces an output no matter how many genes in a gene set are set to 0.
It would be interesting to create a metric(s) in the output data accounting for this (e.g. a simple proportion of 0 genes or a ponderated metric if the misssing genes are the ones with higher weights/importance in the gene set, which could drive to misleading conclusions about the pathway activation). With it the user can get an idea of the "confidence" of the score computation.
This is an example of the current warning of prepareSummarizedExperiment
out <- prepareSummarizedExperiment(input_SE, "gtex_gokegg")
Warning message:
In prepareSummarizedExperiment(input_SE, "gtex_gokegg") :
125 genes present in the model not found in input data. The expression of all samples will be set to 0
When using
prepareSummarizedExperimentthere is a warning informing of the number of genes missing in the input that will be set to 0 for the model. HowevercomputeGeneSetScoresalways produces an output no matter how many genes in a gene set are set to 0.It would be interesting to create a metric(s) in the output data accounting for this (e.g. a simple proportion of 0 genes or a ponderated metric if the misssing genes are the ones with higher weights/importance in the gene set, which could drive to misleading conclusions about the pathway activation). With it the user can get an idea of the "confidence" of the score computation.
This is an example of the current warning of
prepareSummarizedExperimentout <- prepareSummarizedExperiment(input_SE, "gtex_gokegg")