Probabilistic prediction of protein secondary structure using causal networks

In this paper we present a probabilistic approach to analysis and prediction of protein structure. We argue that this approach provides a flexible and convenient mechanism to perform general scientific data analysis in molecular biology. We apply our approach to an important problem in molecular biology - predicting the secondary structure of proteins - and obtain experimental results comparable to several other methods. The causal networks that we use provide a very convenient medium for the scientist to experiment with different empirical models and obtain possibly important insights about the problem being studied.