Circular Dichroism in Protein Analysis
63
The
variable
selection-self-consistent
program SELCON provides analysis based
on a database of 17 proteins. In SELCON,
the proteins in the database are arranged
in an order of increasing root-mean-square
difference from the spectrum; the spec-
tra least like the spectrum of interest
are deleted to increase the speed. Then
the program utilizes the observation that
prediction improves when the protein is
included in the basis set. The initial guess
of the structure of the protein is made,
and the secondary structure of the protein
is then determined. The solution replaces
the initial guess and the process is repeated
until self-consistency is attained. This pro-
gram gives a fairly accurate estimation of
α
-helix,
β
-sheet, and
β
-turn of globular
proteins based on CD spectra between 240
and 200 nm. However, it overestimates
α
-
helix and underestimates
β
-sheet when
polypeptides have high
β
-sheet content.
The neural network program, K2D, uses
a learnt mapping between spectra and sec-
ondary structure content to analyze new
experimental spectra. It was trained on
input data, which were CD spectra rep-
resented as 41 inputs corresponding to
the intensities at the wavelengths between
240 and 200 nm. The output units formed
a self-organized map, which could be
analyzed to give fractional weights of pro-
tein secondary structures. In the training
phase, the weighting of each wavelength
is altered, until the error between the
calculated and actual secondary structure
content for the training data is mini-
mized. For a new spectrum, the data
are fed through the network using the
adjusted weights to give an estimate of
the secondary structure content. The K2D
program predicts the secondary structure
content fairly well, but does not give a
matched spectrum. A different type of neu-
ral network was used, in which the output
units gave directly the fractional secondary
structure content:
α
-helical, parallel and
antiparallel
β
-sheet,
β
-turn, and others.
The network was trained on 13 spectra,
each represented as 83 inputs correspond-
ing to the intensities from 178 to 260 nm.
Each of these methods and programs
has
its
advantages
and
disadvantages.
Even when a method provides a good
match between the calculated spectrum
and the experimental data, it may not
necessarily provide the best estimation
of protein secondary structure content.
Among these methods, SELCON may
perform better than the others. Most of
these programs usually estimate
α
-helical
content well, but underestimate
β
-sheet
and
β
-turn contents.
5
Experimental Aspects
5.1
Instrumentation
Nowadays, measuring CD is routine work
in many laboratories. There are four man-
ufacturers of CD instruments. Aviv and
Associates (Lakewood, New Jersey) has
Cary 61 and 62 CD instruments with a
photoelastic modulator, modernized elec-
tronics, and a computer control system.
JASCO (Easton, Maryland and Tokyo)
makes J710 and J720 spectrometers. Jobin-
Yvon (Longjumeau, France) manufactures
the JY Mark VI and OLIS (On-Line Instru-
ment Systems, Bogart, Georgia) offers the
Cary 61 and 62 CD instruments. CD ac-
cessories for stopped-flow instruments are
available from Applied Photophysics Ltd.
(Leatherhead, UK) and Biologic (Greno-
ble, France).
Most instruments use a single beam.
In Eq. (2), the intensity of the incident
previous page 1383 Encyclopedia of Molecular Cell Biology and Molecular Medicine read online next page 1385 Encyclopedia of Molecular Cell Biology and Molecular Medicine read online Home Toggle text on/off