GENEOnet: A New Machine Learning Paradigm for Protein Pocket Detection Based on Group Equivariant Non-expansive Operators

Artificial intelligence (AI) is becoming more and more pervasive in our daily lives, and in particular, it often consists of deep learning models. Deep learning continues to prove its effectiveness in a wide range of applications, but its lack of transparency is also well known: its decisions are not always explainable, even if every stage of the model is fully known. Indeed the transparency of these models usually decreases as their complexity increases. Hence the interest in building simple AI structures obtained by injecting knowledge in the learning system, exploiting the properties of Group Equivariant Non Expansive Operators (GENEOs). Here a successful case study of application of GENEOs to protein pocket detection is presented.