How to interpret similar motifs in Homer results

Question

[Aspire]

Do several similar motifs add to the strength of evidence, or is the information from similar motifs redundant?

For example, does the similarity of motifs 1 & 3 (from Homer2) strengthen the evidence for each ?

Also, in the image above, motif #4 is the reverse complement of motifs 1&3. Is that additional evidence, or mere redundancy ?

[Mensur Dlakic]

You probably already know this, but just in case. Motif #2 is completely different from the rest, and not just because it is longer. It is likely bound by a homodimer (it is a pseudo-palindrome) and has that characteristic dip in base conservation in the middle of it.

Answer 1

[Mensur Dlakic]

The concept of proteins binding a given strand of DNA is our own mental concoction, and doesn't match reality. Any binding site that is at least 5-6 bases long will have the first and last base rotationally separated by at least 180 degrees because of DNA twisting. Given that most proteins don't wrap around DNA and instead bind it it from one side, it is almost a given that they will interact with bases and phosphates from both strands. For our convenience we usually represent binding sites so they match whatever is the upper strand with regard to transcription that is regulated from that binding site. There are two problems with that: 1) we don't always know whether the binding site in a converging intergenic region regulates the top-strand gene going to the right, or a bottom-strand gene going to the left; 2) not everyone designates the binding site strand with regard to downstream genes.

This is a long-winded way of saying that your motif #4 is almost guaranteed to be the same thing as motif #3. They are so perfectly symmetrical when invoking reverse complementarity that it would difficult to explain it otherwise.

Answer 2

[Alex Reynolds]

There is a fair amount of redundancy in motif models. You might take a look at Jeff Vierstra's work on clustering motifs by similarity measures, or run your motifs through TOMTOM to perhaps do pairwise measurements.

The presence of a reverse-complement motif model would indicate the availability of the complementary sequence to a transcription factor, although the way that DNA unfolds and binding occurs is not necessarily symmetric. TOMTOM should take into account orientation when measuring similarity.

ChIP-seq experiments should help provide evidence of binding of any particular TF, which may help with interpretation.