Reasoning behind TomTom defaults on MEME server

[Spencer Nystrom]

Since the defaults for TomTom vary between the commandline & server versions I was wondering whether the server defaults should be preferred if running TomTom on DNA motifs? Related, is there new information or analysis that has been done which changes the conclusions from the original paper? For example, the TomTom paper suggests euclidean distance performs best, but the default is now pearson.

Thanks,

  -Spencer

[cegrant]

Based on the original Tomtom paper we probably should have made ED the default comparison scheme.

However, a second paper came out in 2011 which introduced an improved scorings scheme for Tomtom:

Emi Tanaka, Timothy L. Bailey, Charles E. Grant, William S. Noble and Uri Keich, "Improved similarity scores for comparing motifs" Bioinformatics 27(12): 1603-1609, 2011

under the improved scoring scheme, the difference between the various comparison functions is relatively minor. We've left the Pearson CC as the default, so that folks repeating analyses wouldn't be puzzled by their results changing.

[Spencer Nystrom]

Ah, I see. I for some reason had never seen that second paper.

To clarify, did the 2011 paper result in the `-incomplete-scores` flag? And since then TomTom uses complete scores by default?