We extend PRIME, an intermediate-resolution protein model previously applied to polyalanine and polyglutamine, to the description of the geometry and energetics of peptides containing all twenty amino acid residues. The PRIME model has been used previously in conjunction with discontinuous molecular dynamics, to simulate the self assembly of large systems of polyalanine peptides into fibrillar structures. The goal here is to develop a similar model but now applicable to all twenty amino acids, which would allow us to simulate the spontaneous fibrillization of heterogeneous proteins, including the Alzheimer's peptide A-beta. The parameters for extended PRIME include hydrogen-bonding , side-chain/side chain and side-chain/ main chain interaction range and strength, excluded volume parameters and side-chain bond lengths. These are obtained by applying a learning algorithm and linear programming technique that compares the known native states for a selection of proteins from the PDB with decoy structures generated by Levitt. The interactions are fine-tuned by seeking the optimal energy gap between native and decoy structures. Simulations on systems containing peptides known to form fibrils are being performed using the new parameter set.