Particle clustering in turbulent flows is ubiquitous in particle-fluid two-phase flows in the riser of fast fluidized beds. Controlling the level of clustering is very crucial for improving the reactive efficiency in such flows. The present paper is devoted to numerically studying particle clustering in turbulent flows, where both point particles and finite-size particles are taken into account, respectively. A pseudospectral method with the particle tracking method is used to simulate the point particle clustering in the forced isotropic turbulent flows, while the lattice Boltzmann method is used to simulate the finite-size particles in decaying isotropic turbulent flows. It is observed that the mean drift velocity, caused by forces such as gravity, could reduce the level of particle clustering for point particles. However, the particle clustering is very weak for the particles of sizes larger than the characteristic viscous scales of turbulent flows, which is consistent with the previous experimental results. The observations suggest that the mean drift velocity and particle sizes can be used to control the level of particle clustering in chemical reactors.