This project is an example how the recent advancements in Neural Networks can be applied to sequence data, namely DNA.

We will take a collection of sequences plus information if they are bound by a transcription factor as input data and after training a convolutional neural network we will be able to make predictions for new sequences. In addition we will extract what the network learned and make a plot of the motif.

The example are chosen such that it is not necessary to have a GPU and should learn just fine on a CPU.

To make sure everybody can play around with the example easily and you do not need to install the dependencies, please follow the instructions below.

The code for this tutorial is written in Python and you will need a 2.7 or 3.5 installation (https://www.python.org/downloads/). If you have

Neural Networks are a complex topic and there are quite a few packages you need to install to get going. The easiest way to install packages in Python is to use Anaconda. In the following I will assume that you only have Anaconda installed.

To run the the code in the tutorial you will need the following libraries:

• Jupyter / IPython
• Keras
• Theano or Tensorflow
• Numpy

After the successful installation of Anaconda we will create a new conda environment to not pollute the default environment:

$ conda create -n seqnn python=3.5 

# On Windows
$ activate seqnn
# On Mac / Linux
$ source activate seqnn

Install Theano

$ conda install theano pygpu

Install other dependencies

$ conda install scikit-learn keras numpy scipy matplotlib ipython jupyter pandas sympy nose nb_conda

Download the Neural_Network_DNA_Demo project either by cloning it with git

$ git clone https://github.com/const-ae/Neural_Network_DNA_Demo.git

OR download this zip file and extracting it somewhere.

Move with the command line to the Neural_Network_DNA_Demo folder:

$ cd <LOCATION_OF_DOWNLOAD>/Neural_Network_DNA_Demo

and start the Jupyter notebook:

$ jupyter notebook

A new browser page should open where you can click on the .ipynb file and start experimenting.

If you have problems with the installation of ...

• ... Theano --> check this guide
• ... Keras --> check this guide

or just open an issue here.