SSNN, a method for neural network protein secondary structure fitting using circular dichroism data

Abstract

Circular dichroism (CD) spectroscopy is a quick method for measuring data that can be used to determine the average secondary structures of proteins, probe their interactions with their environment, and aid in drug discovery. This paper describes the operation and testing of a self-organising map (SOM) structure-fitting methodology named Secondary Structure Neural Network (SSNN), which is a methodology for estimating protein secondary structure from CD spectra of unknown proteins using CD spectra of proteins with known X-ray structures. SSNN comes in two standalone MATLAB applications for estimating unknown proteins' structures, one that uses a pre-trained map and one that begins by training the SOM with a reference set of the user's choice. These are available at http://www2.warwick.ac.uk/fac/sci/chemistry/research/arodger/arodgergroup/research_intro/instrumentation/ssnn/ as SSNNGUI and SSNN1_2 respectively. They are available for both Macintosh and Windows formats with two reference sets: one obtained from the CDPro website, referred to as CDDATA.48 which has 48 protein spectra and structures, and one with 53 proteins (CDDATA.48 with 5 additional spectra). Here we compare SSNN with CDSSTR, a widely-used secondary structure methodology, and describe how to use the standalone SSNN applications. Current input format is Δε per amino acid residue from 240 nm to 190 nm in 1 nm steps for the known and unknown proteins and a vector summarising the secondary structure elements of the known proteins. The format is readily modified to include input data with e.g. extended wavelength ranges or different assignment of secondary structures.