...tures ofsproteins.sEntropy is a measure of the information needed to describe a random variable 47. Thesentropy H(X) of a discrete random variable X can be calculated as:s∑ Ω∈ −= x xpxpXH )(log)()( 2s=-=(1)-=-swhere Ω is the set of allowed states, x is an element of Ω, p(x) is the probability of the state x.sThe mutual information I(X; Y) is a measure of correlation between two discrete randomsvariables, X...

...he state x.sThe mutual information I(X; Y) is a measure of correlation between two discrete randomsvariables, X and Y, which is defined as:s7 ∑= −= yx ypxp yxpyxp YXHXHYXI , 2 )()( ),(log),( )|()();(s=-=(2)-=-swhere p(x ,y) is the joint probability of observing states x and y. If the random variables aresindependent then the mutual information achieves its lowest bound of zero. Mutualsinformation cannot ex...

...3615. However, the running time is increasedssignificantly. While our DW structure alphabet is also developed in Cartesian coordinatesspace with the same fragment length and similar number of letters =-=(28)-=-. Our dual-layer modelsgets Q-score of 0.456, which is higher than the best results of Sander et al. by nearly tenspercent. Such huge difference may be caused by a variant of factors. One important fa...

...entropy of either variable.sIn this study, all the probabilities are calculated from the training set by the maximalslikelihood estimation. The probability p(x) of state x is defined as:sNnxp x /)( =s=-=(3)-=-swhere nx is the number of observations of state x, N is the total number of samples. Nossmoothing is used since the training set is sufficiently large.sDual-layer predictionsThe correlations between ...

...t toolsfor focusing on key regions with high prediction accuracy. There are many definitions of thesRI. Here, we use the definition proposed by Rost and Sander 33:s))(0INTERGER(1 max nextoutoutRI −×=s=-=(4)-=-s10 where outmax is the output of the unit with highest value and outnext is that of the unit with thesnext highest value. The factor 10 normalizes RI to integer values from 0 to 9.sResults and Discus...