Quick Start Guide

See installation for directions on downloading and installing DeFCoM.

Running DeFCoM

Once DeFCoM is installed, two executable scripts should be callable from a command line interface train.py and predict.py. Both require a DeFCoM configuration file as a command line argument. An example configuration file is provided in the package directory at defcom/data/example.cfg which can be alternatively found here. This should be used as a base template and modified for your data. Details on what each parameter represents can be found in the configuration file section of this user guide.

Given a config file named example.cfg, train a model using the command

train.py example.cfg

After the model is trained, you can predict footprints with the command

predict.py example.cfg

If train.py and predict.py are not recognized as exectuables by your system, these can alternatively be found in the bin directory of the DeFCoM package and can be run on the command-line with the command structure

python /path/to/bin/script_name.py path/to/config/config_file.cfg

DeFCoM Output

When finished running, DeFCoM will provide results in a BED format like file that contains scores (last column) for each of the motif sites input during DeFCoM’s classification phase. For true binary classifcation, a score of 0 can be used as a cutoff with scores above 0 indicating TF-bound sites, though we do not recommend applying results this way as it will likely generate a high type I error rate (false positives). Instead, we suggest applying the scores as a ranking mechanism with higher scores indicating a stronger confidence in a motif site being TF-bound.