Cognitive radio (CR) is of crucial importance in providing efficient management of limited spectrum resources. However, its performance relies on efficient spectrum sensing. This paper investigates a novel approach for CR networks that leverages intelligent reflecting surface (IRS) specifically for spectrum sensing and non-orthogonal multiple access (NOMA) for data transmission. We propose a Grey-Wolf Optimization (GWO) based IRS optimization approach to maximize spectrum sensing performance. Independent of the IRS, NOMA is employed to improve spectral efficiency during data transmission. The performance is evaluated in terms of throughput and spectrum sensing parameters, namely probability of false alarm and missed detection. Numerical and simulation results demonstrate that GWO-based IRS optimization significantly outperforms conventional nature-inspired algorithms, achieving approximately 97% improvement in spectrum sensing accuracy. Based on the improved spectrum sensing results, the effective data transmission throughput is evaluated and validated through extensive simulation.