An approach to the solution of large-scale nonlinear programming problems with initeger restrictions on some of the variables is described. The method is based on the MINOS large-scale optimization algorithm and involves adjustment of the superbasic variables in integer steps. The directhsearch procedure commences from a continuous (relaxed) optimal solution and seeks the nearest integer-feasible solution in terms of objective function value. Computational experience on a wide variety of nonlinear problems is reported.