Comparing motifs from TFMoDisco and other methods?

[sandraleebatista]

What are recommended ways to compare motifs from TFMoDisco and other methods in terms of similarity and significance?

[akundaje]

MODISCO produces 3 versions (representations) for each motif it learns - PFM (position frequency matrix) - CWM (contribution weight matrix) - hCWM (hypothetical contribution weight matrix) You can get a more intuitive explanation of these in the BPNet paper https://www.nature.com/articles/s41588-021-00782-6 The PFM versions of each motif are classic motifs capturing frequencies of bases at each position. So you can easily compare these to other motifs if you like using standard motif comparison tools. We prefer using the CWM representation as the primary motif representation since it poses several advantages over PFMs. Again see the BPNet paper for more details.

…

On Tue, Jun 21, 2022 at 12:24 PM sandraleebatista ***@***.***> wrote: What are recommended ways to compare motifs from TFMoDisco and other methods in terms of similarity and significance? — Reply to this email directly, view it on GitHub <#103>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AABDWEM6AGGFZPDPR4HPPCDVQIJGHANCNFSM5ZNPTPLQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>