Evaluating Hyperparameter Optimization in Machine Learning Algorithms for Cancer Driver Gene Classification
Resumo
Cancer driver genes (CDGs) play a central role in tumorigenesis and represent important targets for diagnosis and therapy. In this study, we evaluate the impact of hyperparameter optimization on the predictive performance of traditional machine learning algorithms using multi-omics data. We perform systematic searches across different configurations to identify the most effective settings for CDG classification. Our results show that optimized models consistently outperform their default counterparts, particularly in recall and precision-recall metrics, with ensemble methods showing the most pronounced gains. These findings indicate that traditional algorithms, when carefully tuned, can represent promising approaches for identifying CDGs from genomics data.
Referências
Schulte-Sasse, R., Budach, S., Hnisz, D., and Marsico, A. (2021). Integration of multiomics data with graph convolutional networks to identify new cancer genes and their associated molecular mechanisms. Nature Machine Intelligence, 3(6):513–526.
Stratton, M. R., Campbell, P. J., and Futreal, P. A. (2009). The cancer genome. Nature, 458(7239):719–724.
Vogelstein, B., Papadopoulos, N., Velculescu, V. E., Zhou, S., Diaz, L. A., and Kinzler, K. W. (2013). Cancer genome landscapes. Science, 339(6127):1546–1558.
