We develop an improved version of PROFESY, a novel method for
ab-initio prediction of protein tertiary structures based on the
fragment assembly and global optimization.
 In contrast to the primitive version presented earlier, where
 hydrogen bond was defined only in terms of inter-atom distance,
 the angle dependence is now correctly incorporated.
 This new feature allows us to obtain low-energy conformations
 with reasonable amount of beta strands, in contrast to earlier
 version where fraction alpha helices was excessively large on average.
  In order to enhance the performance of the prediction method, we
optimize the linear parameters of the energy function, so that the
native-like conformations become energetically more favorable than
the non-native ones for proteins with known structures. We test
the feasibility of the parameter optimization procedure by
applying it to the training set consisting of two proteins of
structural class alpha+beta: 1FSD and 1PQS. We use the
resulting parameter set for the jackknife test, using several
proteins with various structural classes as test sets. The results
are quite promising. In particular, for protein 2GB1, the
prediction results improve dramatically with the optimized the
parameter set compared to the original parameters,  despite the
fact that it is NOT INCLUDED IN THE TRAINING SET. The result
suggests that parameters trained for a relatively small number of
proteins are transferrable to other proteins to some extent.