The ff14SB force field improves the accuracy of protein side chain and backbone parameters from ff99SB. ff99SB, based on ff94, has weaknesses in side chain rotamer and backbone preferences. ff14SB includes a complete refit of all amino acid side chain dihedral parameters, trained on multidimensional dihedral scans for better transferability. Parameters for alternate protonation states of ionizable side chains were also generated. Average relative energy errors between conformations were under 1.0 kcal/mol compared to QM, reduced 35% from ff99SB. Empirical adjustments to backbone dihedral parameters were made, with multiple small adjustments tested against NMR scalar coupling data and secondary structure content for short peptides. The best results were obtained from a physically motivated adjustment to the φ rotational profile, compensating for lack of QM training data in the β-ppII transition region. ff14SB improves secondary structure content in small peptides and reproduces NMR χ₁ scalar coupling measurements for proteins in solution. The Amber ff12SB parameter set, a preliminary version of ff14SB, includes most of its improvements. ff99SB has limitations, including less accurate rotamer preferences for some side chains and issues with secondary structure preferences. ff14SB addresses these by improving side chain dihedral parameters and making empirical adjustments to backbone parameters. The side chain training used complete amino acids in QM calculations, unlike ff99SB which used small organic compounds. The backbone training included multidimensional scans of all side chain χ rotatable bonds with both α and β backbone conformations. The fitting targets were gas-phase ab initio energies. The objective function for parameter optimization was based on relative energy errors between conformations. A genetic algorithm was used to minimize the ANE, perturbing amplitudes and phase shifts. The simulation protocols included equilibration with a Berendsen thermostat and barostat, followed by production simulations with appropriate restraints. ff14SB improves the accuracy of protein simulations, particularly in reproducing NMR scalar couplings and secondary structure content. The parameters for all non-hydrogen dihedrals in the protein side chains are presented in Table S2. The design of the training set included amino acid conformation pairs with simultaneous changes to more than one rotatable bond, necessitating concurrent optimization of parameters for multiple dihedrals. The parameter space for optimization is considerable, but the use of solving groups based on shared dihedral atom types allowed for efficient optimization. The intrinsic backbone dependence (BBD) was calculated to assess the ability of side chain dihedral parameters to match QM data in the absence of explicit coupling between backbone and side chain parameters. The results show that ff14SB improves the accuracy of protein simulations, particularly in reproducing NMR scalar couplings and secondary structure content.The ff14SB force field improves the accuracy of protein side chain and backbone parameters from ff99SB. ff99SB, based on ff94, has weaknesses in side chain rotamer and backbone preferences. ff14SB includes a complete refit of all amino acid side chain dihedral parameters, trained on multidimensional dihedral scans for better transferability. Parameters for alternate protonation states of ionizable side chains were also generated. Average relative energy errors between conformations were under 1.0 kcal/mol compared to QM, reduced 35% from ff99SB. Empirical adjustments to backbone dihedral parameters were made, with multiple small adjustments tested against NMR scalar coupling data and secondary structure content for short peptides. The best results were obtained from a physically motivated adjustment to the φ rotational profile, compensating for lack of QM training data in the β-ppII transition region. ff14SB improves secondary structure content in small peptides and reproduces NMR χ₁ scalar coupling measurements for proteins in solution. The Amber ff12SB parameter set, a preliminary version of ff14SB, includes most of its improvements. ff99SB has limitations, including less accurate rotamer preferences for some side chains and issues with secondary structure preferences. ff14SB addresses these by improving side chain dihedral parameters and making empirical adjustments to backbone parameters. The side chain training used complete amino acids in QM calculations, unlike ff99SB which used small organic compounds. The backbone training included multidimensional scans of all side chain χ rotatable bonds with both α and β backbone conformations. The fitting targets were gas-phase ab initio energies. The objective function for parameter optimization was based on relative energy errors between conformations. A genetic algorithm was used to minimize the ANE, perturbing amplitudes and phase shifts. The simulation protocols included equilibration with a Berendsen thermostat and barostat, followed by production simulations with appropriate restraints. ff14SB improves the accuracy of protein simulations, particularly in reproducing NMR scalar couplings and secondary structure content. The parameters for all non-hydrogen dihedrals in the protein side chains are presented in Table S2. The design of the training set included amino acid conformation pairs with simultaneous changes to more than one rotatable bond, necessitating concurrent optimization of parameters for multiple dihedrals. The parameter space for optimization is considerable, but the use of solving groups based on shared dihedral atom types allowed for efficient optimization. The intrinsic backbone dependence (BBD) was calculated to assess the ability of side chain dihedral parameters to match QM data in the absence of explicit coupling between backbone and side chain parameters. The results show that ff14SB improves the accuracy of protein simulations, particularly in reproducing NMR scalar couplings and secondary structure content.