When the name and the matrix disagree

The factor name attached to a motif is a claim about what binds it, and the claim is only as good as the experiment behind it. Comparing every one of the 4,232 motifs in this atlas against all the others makes three kinds of disagreement visible. None of them means a database is wrong — a ChIP experiment faithfully reports what was crosslinked, which is not always the tagged protein.

Closest relatives contradict the label

53 motifs whose nearest matches across all four databases belong to a different family than their own name, while nothing in their own family comes close. Frequently these are composite elements or indirect binding: a factor recruited to another factor's site reports that site's motif. The 22 that cross a structural class are listed first and are the interesting ones; the rest are mostly homeodomain matrices sitting between two specificity classes, which the atlas splits on the single residue at homeodomain position 50.

Motifs inherited from another species

83 CIS-BP entries carry a human gene name but were measured on a different organism's protein, and transferred on DNA-binding-domain similarity. The weakest transfers are listed first; a value of 1.00 means an identical DBD.

Uncharacterised motifs that match a known family

68 motifs from the zinc-finger buckets — where the label says little more than "a zinc finger" — are essentially the motif of a well-defined family. Read these as a lead, not a conclusion: two proteins can share a site.