Hi,
I'm using miRDeep to discover new miRNA in Illumina HTS data. I don't understand the -v and the -z parameters.
here's the definition in the readme of the script
-v (for instance, -v 5) varies the cut-off. This is the single most important and difficult option. In
general, if you are missing a large fraction of known miRNAs that you know are in the dataset, you should
lower the cut-off. If you make the permuted controls (see below) and get many false positives, you should
raise the cut-off. As a rule of thumb, the cut-off should probably not be lowered below 0 or 1 unless the
dataset is _highly_ enriched in miRNA sequences.
the definiton is ok, but how to set a good thresold ? on what is it filtering ?
-z scores the stems below the potential precursor sequence for potential Drosha recognition. In general,
more base pairings between the 10 nt flanking the potential precursor suggest a tighter stem and gives a
higher score contribution. For each number of base pairings b in the stem, the probability p(b) of a
miRNA precursor hairpin or a background hairpin having p(b) was estimated, and log-odd score contributions
were calculated. When this option is used, we recommend that a slightly higher cut-off is used (as default
1.5).
can someone explain me that in other words, I'm a little lost there.
Thanks
I'm using miRDeep to discover new miRNA in Illumina HTS data. I don't understand the -v and the -z parameters.
here's the definition in the readme of the script
-v (for instance, -v 5) varies the cut-off. This is the single most important and difficult option. In
general, if you are missing a large fraction of known miRNAs that you know are in the dataset, you should
lower the cut-off. If you make the permuted controls (see below) and get many false positives, you should
raise the cut-off. As a rule of thumb, the cut-off should probably not be lowered below 0 or 1 unless the
dataset is _highly_ enriched in miRNA sequences.
the definiton is ok, but how to set a good thresold ? on what is it filtering ?
-z scores the stems below the potential precursor sequence for potential Drosha recognition. In general,
more base pairings between the 10 nt flanking the potential precursor suggest a tighter stem and gives a
higher score contribution. For each number of base pairings b in the stem, the probability p(b) of a
miRNA precursor hairpin or a background hairpin having p(b) was estimated, and log-odd score contributions
were calculated. When this option is used, we recommend that a slightly higher cut-off is used (as default
1.5).
can someone explain me that in other words, I'm a little lost there.
Thanks
Comment