Inertial particle acceleration statistics are analyzed using DNS in the case of a highly turbulent channel flow. Along with effects recognized in homogeneous isotropic turbulence, an additional effect is observed due to high and low speed vortical structures aligned with the channel wall. In response to those structures, the inertial particles experience strong longitudinalacceleration variations. DNS is also used in order to assess LES-SSAM (Subgrid Stochastic Acceleration Model), in which an approximation to the instantaneous non-filtered velocity field is given by simulation of both, filtered and residual, accelerations. Advantages of this approach in predicting particle dynamics in the channel flow at a high Reynolds number are shown.