protein sequence training

hi All :)

i have protein sequences data, then i wanna to train it with neural network backpropagation, but i have a problem with data training.. because,the number of input unit and output unit are different and heterogeneus, how to equalize the number of input unit and output unit?? for example input: HKJMMLLKNJHVHVHCVGVJKLHCVGCGCH output: JMMLL

input: fdsfnhldglsfhldsfhidjgfodjgijidshfjzhnjfndjngvfngidjgxbvfgd output: idjg

there are poeple who suggest to use sliding window, but i'm so confused how to implement it

thx for ur suggestion before :)

[Chandra Kurniawan]

Hi,miftahurrahma.

Are you sure you want to treat HKJMMLLKNJHVHVHCVGVJKLHCVGCGCH as input of neural network?

As we know that neural network takes numerical input.

[Greg Heath]

What do the input and output sequences represent?

Answer 1

[Greg Heath]

In general, the number of elements in the input vector, I, and the number of elements in the corresponding output vector, O, are different.

Decide what the output sequence should be given the corresponding input sequence. Then encode each so that it is a column vector of length I and O, respectively.

Finally form input and output matrices from N i/o pairs to have dimensions

[ I N ] = size(inputmatrix)

[ O N ] = size(outputmatrix)

Hope this helps.

Greg

Answer 2

[miftahurrahma rosyda]

i will encode that sequence with its protein properties, so, the input is numeric..

Answer 3

[lei Xiao]

It is a complex biology process and need professional technology to support you. It would be helpful to visit http://www.creative-proteomics.com/about-us.htm