Basically, the num of luminescence is caused by the spatial structure of the certain protein. However, environmental factors
are inevitable and these factors also influence the total number of luminescence.
We assume the environmental factors' influence can be replaced by the random change
of the sptial structure.
Assume x is the relative position of the sequence, and s(x) is the score of each
position. Including the random factor, we have
Luminescence is a function of s*(x), assume the function is f, so
We use a multi-layer BP neural network to fit funtion g.
We set the number of neuron in the output layer to one, and use Relu function to calculate g(x1,x2, ... ,xn). Relu is defined
as following:
By this network, we succesafully fitted g(x1,x2, ... ,xn). Then we put the sequence of OR1D2 as the input, with the mentioned
network, we can predict the total number of luminescence of OR1D2 in the same( or
similar) environment.
