K2d Frequently Asked Questions

M.A. Andrade, P. Chacón, J.J. Merelo

1.- What are the circular dichroism units in k2d?

The typically used units for molar ellipticity ( [

] ) in biochemistry are:

degrees cm²/dmol

However, molar circular dichroism: (

) is quite used as well. The conversion between the two measures is given by: [

] = 3298

. For more details about CD see [1].

The k2d input should be your CD spectrum in molar ellipticity multiplied by 0.001. These are the units of the training set we took [2]. Then, if your measures are in molar circular dichroism units you have to multiply your spectrum by a factor of 3.298.

2.- Will be more versions of k2d?

Yes. There is a new version: K2D2. The update is explained in this manuscript:
Perez-Iratxeta, C. and M.A. Andrade-Navarro. 2008. K2D2: Estimation of protein secondary structure from CD spectra. BMC Structural Biology. In press.

3.- Do you know a protein CD database in the internet?

We know one, kindly provided by Petr Pancoska at the University of Illinois at Chicago. You can access there (Electronic) CD spectra of 28 proteins in a two column format with wavelength and intensity (units:

10^5) [3].

4.- Which are the papers in reference to k2d?

There are strictly two. In the first we introduced the network and optimized its parameters [4]. In the second we developed the network evaluation method comparing it with the classical ones [5].
We have been, very positively, mentioned in a review [6].
The update of k2d is presented in [7].

5.- Information about CD on the WWW:

Circular dichroism spectroscopy A chapter of the Protein Secondary Structure course by Kurt D. Berndt, of the Karolinska Institute, Stockholm, Sweden, for the Internet Course on The Principles of Protein Structure organised by Birkbeck College.
Lectures in Optical rotation and circular dichroism Spectra by Allen Bush at the University of Maryland Baltimore County
Review on CD for studying protein conformation by T. Bae, College of Biological Sciences - University of Minnesota

6.- Which proteins were used to train the neural network?

We used 18 proteins: myoglobin, lactate dehydrogenase, lysozime, cytochrome C, subtilisin BPN, papain, ribonuclease A, alpha-Chymotrypsin, elastase, concanavalin A, parvalbumin, adenylate kinase, insulin, carboxypeptidase A, thermolysin, trypsin inhibitor, ribonuclease S and nuclease.

7.- My spectrum looks OK but k2d cannot find a solution. Why?

The algorithm was trained with a limited set of globular proteins. It is able to interpolate between the learned examples. If your spectra cannot be interpolated the algorithm will give you probably the interpolation that is closest to your spectrum but noting that the solution should not be accurate. In such a case, k2d is just not the appropriate algorithm to apply and you may try another one (check the review from Greenfield [6]).

8.- k2d handles the range 200-240 nm but I am interested in the 190-200 nm region. Do you have another version of k2d or do you know of any other software which can handle my data?

It is our intention to develop a new version that extends the range to 190 but it is still not ready. Meanwhile you can use other soft (see a short revision by T. Bae). We recommend CCA or SELCON. You can obtain those programs and others from N.J. Greenfield (for details see [6]).

9.- What does the square distance value represent?

The neural network tries to fit your spectrum to a combination of the spectra it was trained with. Indeed, it always makes one that is the closest one to yours (the best it can do). However, a last quality measurement is to check how similar is this interpolated spectrum to the spectrum you provide. This is measured with the distance between the two spectra (a distance between two vectors). From the analysis of known data we know that distances longer than 0.227 mean that the result given by the network is not reliable and should be discarded.

References:
[1] Woody, R.W. 1995. Methods in Enzymology, 246, 34-71
[2] Yang, J.T., C.C. Wu and H.M. Martinez. 1986. Methods in Enzymology, 130, 208-271
[3] Pancoska, P., E. Bitto, V. Janota, M. Urbanova, V.P. Gupta and T.A. Keiderling. Prot. Sci., 4, 1384-1401
[4] Merelo, J.J., M.A. Andrade, A. Prieto and F. Morán. 1994. Proteinotopic Feature Maps. Neurocomputing. 6, 443-454
[5] Andrade, M.A., P. Chacón, J.J. Merelo and F. Morán. 1993. Prot. Eng., 6, 383-390 Medline
[6] Greenfield, N.J. 1996. Anal. Biochem., 235, 1-10
[7] Perez-Iratxeta, C. and M.A. Andrade-Navarro. 2008. K2D2: Estimation of protein secondary structure from CD spectra. BMC Structural Biology. In press.

This page uses a GIF math symbol collection created by Karen M. Strom