A Collaborative Approach to Improving Missense Mutational Effect Predictions in Oncoproteins

The current paradigm for predicting the effects of missense mutations relies on a variety of computational tools, each with its own strengths and weaknesses. This paper operates within this established paradigm but seeks to refine it.

This preprint presents the "APE score," a novel ensemble metric that combines the predictions of three existing unsupervised methods (AlphaMissense, PRESCOTT, and ESM1b) for evaluating the impact of missense mutations, specifically in oncoproteins. It demonstrates improved performance over individual component methods and establishes classification thresholds for pathogenicity. The work represents a refinement of existing tools within the current paradigm of missense mutation effect prediction and does not propose a radical shift in methodology or understanding. Therefore, it fits squarely within Kuhn's definition of "normal science" – puzzle-solving activity within an accepted paradigm. The authors themselves acknowledge the iterative nature of their work within the context of the ongoing progress in the field. This continuous improvement and refinement of prediction methods is characteristic of normal science, as researchers strive to improve accuracy and applicability within the existing framework.