AvantiShri

#install keras from https://github.com/kundajelab/keras/tree/keras_1
from __future__ import print_function
import keras
import numpy as np
np.random.seed(1)

#build a sample model
model = keras.models.Sequential()
model.add(keras.layers.convolutional.RevCompConv1D(input_shape=(100,4),
                                                   nb_filter=10,

AvantiShri

AvantiShri

AvantiShri

AvantiShri

"""
Author: Avanti Shrikumar

Here's a gist containing code to run in-silico mutagenesis (ISM)
on a model that takes one-hot encoded DNA sequence as the input. The
ISM score at a base is the prediction when that base is
present minus the average prediction across all 4 possible bases at
at that position.

"prediction_func" needs to be a function that maps one-hot encoded sequence