A simple and efficient approach to the protein secondary structure prediction problem is presented and evaluated with four established measures: Q3, Matthews coefficients, Qobserved and Qpredicted. They are applied to the raw data and also to features extracted with the PCA and the ICA methods. The results obtained are better than any predictor trained in similar conditions.
Published in:
Electronics Letters
(Volume:40
,
Issue:
21
)
Date of Publication: 14 Oct. 2004