The genomes of organisms have been sequenced for many years, leading to the discovery of thousands of genes. DNA microarrays are widely used tools for simultaneously analysing numerous genes, commonly employed in detecting and identifying various diseases, including cancer. However, microarray datasets have non-relevant and redundant information, hindering their analysis. This problem is further exacerbated considering the datasets’ high dimensionality and imbalanced classes. Consequently, standard practice involves incorporating a feature selection process to identify the most relevant genes and their associations with diseases. Various methods have been employed to address this task. However, none have taken a more holistic approach that effectively handles feature selection, automatically identifies the optimal classifier configuration, and manages potential conflicting objectives simultaneously. In response, this study introduces the -metric selection - multiobjective neuroevolution of augmenting topologies (SMS-MONEAT) algorithm, which combines the multiobjective optimisation framework from -metric selection - evolutionary multiobjective algorithm (SMS-EMOA) and the evolutionary operators from the neuroevolution algorithm N3O, a variation from NEAT which stands for ‘3 new operators’. SMS-MONEAT algorithm was designed to perform both feature selection and optimise the configuration of artificial neural networks for classification tasks. SMS-MONEAT was evaluated against classic and state-of-the-art methods for feature selection and microarray classification. The experiments were conducted on 20 highly challenging cancer-type datasets primarily sourced from the Curated Microarray Database, and the results were investigated for statistical significance. The findings suggest that SMS-MONEAT either outperforms or achieves competitive results in terms of classification compared to the mentioned methods, while at the same time, it selects a smaller subset of features.