# How do I computationally predict the binding position of a DNA binding protein on the promoters of differentially expressed genes?

I have a list of human genes that have been proven to be up-regulated in a disease condition of interest, using microarray analysis. I also have a protein which according to literature is said to be a DNA binding protein. I wish to be able to computationally predict if the protein mentioned binds to any of the promoter sequences of my differentially expressed genes and if yes , then at which position?

Are there any Bioinformatic tools available for this task?

You could run FIMO on your entire genome for TFs (transcription factors; DNA-binding proteins) of interest, which gives you binding sites: genomic intervals where those TFs bind.

https://bioinformatics.stackexchange.com/a/2491/776

You can then map or intersect promoter regions with those TF binding sites with BEDOPS bedmap or bedops, respectively:

The use of bedmap is recommended to assign TFs to a promoter of interest. This tool will return both the promoter region and any TFs that overlap it by the specified overlap criteria.

• Thanks a lot for your very helpful response. "FIMO scans a set of sequences for individual matches to each of the motifs you provide ". So how do I relate ouput from FIMO to TF binding sites and the genomic intervals? Is there any method to get the specific names of the genes whose promoter regions bind my TF? I have used the FIMO web tool meme-suite.org/tools/fimo. May 31 '19 at 8:50
• I don't understand. If you have promoters, those are derived from genes, no? May 31 '19 at 12:20
• Sorry for the silly question, but which column of FIMO output corresponds to the TF binding sites? I now want to intersect promoter regions with those TF binding sites using BEDOPS bedops. May 31 '19 at 14:03
• The output from the command-line tool should be in BED format, so the binding site of the TF would be the first three columns. I don't know what format comes out of the web tool, but I would start by reading the documentation. May 31 '19 at 14:31
• It seems like you might be doing things out of order: You would feed the whole genome (mappable regions) into FIMO against Jaspar or other published motif databases, limiting your search to motifs no longer than, say, 20nt. Then do your set operations against this result with your promoters, at a desired statistical threshold (1e-5 or less). If you're looking to predict putative motifs from your promoters directly, then you would use MEME, not FIMO. Try re-reading the answer I gave. Jun 5 '19 at 0:38

If you work with human, I would start by intersecting your promoter coordinates with the ReMap2018 database which is a comprehensive collection of published ChIP-seq datasets. Extract those TFs that fit your research question. Alternatively you can browse NCBI for published datasets of your TF and then intersect those with your annotations.

Also cross-posted: https://www.biostars.org/p/381751/

• thanks a lot for your response but can you please elaborate a little more? For e.g. I do not understand the meaning of " intersecting promoter coordinates" with a data base. Nov 17 '19 at 11:25
• The database provides lists in BED formats and you have genomic coordinates. You something like bedtools intersect to intersect them to scan for overlaps. Dec 17 '19 at 12:14