search

rajanil / msCentipede

A hierarchical multiscale model for inferring transcription factor binding from chromatin accessibility data.
MIT License
25 stars 6 forks source link

FIMO scans #6

Closed daquang closed 9 years ago

daquang commented 9 years ago

Do you have any tips on using FIMO for scanning? I'm doing de novo motif discovery on open chromatin, and I know msCentipede requires being a good number of true binding sites, but it's hard to know whether I actually am feeding it binding sites without any ChIP data. Got any tips?

rajanil commented 9 years ago

In the paper describing msCentipede, we didn't use FIMO to select putative binding sites. We selected motif sites for which the ratio of likelihood of the position weight matrix for the transcription factor to the likelihood of a uniform background matrix is high (> 1000). You could use a simple peak caller on the ATAC-seq / DNase-seq data, and ensure that a reasonable fraction of motif sites in the training set fall within these peaks.