First, please check PDNAsite server (https://www.nature.com/articles/srep27653). It can support you to know which parts of a protein strongly have a tendency to interact with the target DNA molecules. Similarly, another option for you is StackDPPred server (https://bmll.cs.uno.edu/add) that freely available for academic users.
Next, DNAbinder (http://crdd.osdd.net/raghava/dnabinder/), DP-bind (http://lcg.rit.albany.edu/dp-bind/) and some other servers or offline tools can also provide a user-friendly platform to handle your project. More importantly, there are some python scripts in the literature for prediction the position of DNA-binding proteins.
If I was true, I think Rostlab in Munich recently developed a python script for prediction of DNA-binding proteins. You can check their homepage or their page in GitHub for finding that script I discussed. Their script can predict the position of DNA-binding proteins through amino acid sequence.
Before choosing software in this field, you should be assuring about the sensitivity of target tools for prediction of DNA-binding residues. All in all, the sensitivity of applied algorithm in such servers or tools must be over %60, respectively. By and large, I think you should try to delve the literature to find other suitable tools for your purpose.
Yes, HOMER is an excellent tools for such tasks. Here you can find a nice tutorial:
http://homer.ucsd.edu/homer/ngs/peakMotifs.html
This will show you how to carry out a so-called de novo motif analysis and find the most frequent motifs under your sequences of interest in an unbiased manner (no database used prior to searching - only at the end for comparison).