K2d home page


Welcome to the k2d home page. Here we offer you an algorithm for the estimation of the percentages of protein secondary structure from UV circular dichroism spectra using a Kohonen neural network with a 2-dimensional output layer. You can either use k2d via a web server or get the program and run it on your machine.

We have used a self-organising neural network to extract from a set of circular dichroism spectra ranging from 200 nm to 241 nm the secondary structure features present in the data.

This information is kept in a matrix result of hundred different trainings of the neural network. k2d use this pre-calculated data to give you the result upon your CD data. The run does not take more than a few seconds and provides you with an estimation of the percentages of helix, sheet and random structure of your protein.

k2d also gives the probable error in the estimation, based in the training procedure results.


The Server

You only need to submit 41 CD values ranging from 200 nm to 240 nm (given in deg cm^2 dmol^-1 multiplied by 0.001) and the server gives back the estimated percentages of helix, beta and rest of secondary structure of your protein plus an estimation of the accuracy of the prediction.


The Program

k2d is written in C. We offer the source code and the executable files for either Unix or PC, together with a weight matrix file, an example file and the instructions for its use.

Instructions for k2d

See the instructions for k2d, as they are shown in the k2d.read.me file.


Get the program

You can where these files are.

Alternative access to the ftp directory

ftp anonymous to swift.embl-heidelberg.de directory /group/andrade


Frequently Asked Questions about k2d


Cite us!

The followings publications are in reference to this work:


Please send comments or suggestions to andrade@embl-heidelberg.de
[Sander Group] [Andrade Home]