Train data on new real TFBS scores

[iamciera]

Experiment

Try training network with 4 new TFBS (hb, kni, kr, twi) scores. The news scores are located here: https://drive.google.com/open?id=1PTmQwwgBzEDFNDVRdGMS8bn5gZgk8jJt

Hypothesis

The TFBS new scores are informative of enhancer identify, therefore, including them should increase the accuracy of the model. Alternative results could be the accuracy does not change, in that case it could mean two things these new scores are not informative.

This experiment will be informative along with another experiment in which we add "random" PWM scores. Further, we can start asking which TFBS are the most informative by isolating and "scrambling" only certain score columns.