Score and matching results

[hirokohmatsui]

Hello author,

Thank you for providing the great tool!

I have a question about matchMotifs() results.
When I run the function with the different sets of GRanges (as the second argument) but some are overlapping, I get different results for the exact same ranges.
For example, when I run with GRanges [peak1, peak2, peak3], all peaks matched to the motif. However with GRanges [peak1, peak2, peak3, peak4], peak1 doesn't match anymore.

I assume that the matching scores change depending on the number of GRanges input, so the matching results change as well.
I thought the matching scores are purely defined based on the genomic sequence comparison with some bias, but was I wrong?
Is there any written explanation how you score and set the threshold for the matching?

Best,
Hiroko

[AliciaSchep]

@param bg background nucleotide frequencies. Default is to compute based on
subject, i.e. the specific set of sequences being evaluated. See Details.

#' @details Background nucleotide frequencies can be set to "subject" to use the
#' subject sequences or ranges for computing the nucleotide frequencies,
#' "genome" for using the genomice frequencies (in which case a genome must be
#' specified), "even" for using 0.25 for each base, or a numeric vector with A,
#' C, G, and T frequencies.

By default the input sequences are used to compute background nucleotide frequencies. This isn't great for very small input datasets and could explain different results with 4 vs 3 peaks. Try setting the same background frequencies by specifying the genome or a set vector of frequencies.

[hirokohmatsui]

Hi AliciaSchep,

Thank you for the very clear explanation and quick response.
I appreciate your help!
Hiroko