A Transfer Learning Exploited for Indexing Protein Structures from 3D Point Clouds

摘要

In this paper, we propose a transfer learning-based methodology that can be exploited for indexing protein structures from associated 3D point clouds. Such a methodology can be particularly useful for biologists that are searching automated solutions to find family members of a query protein or even to label new structures by directly using input raw 3D point clouds. Comparative study and performance evaluation show the efficiency and the potential of the proposed methodology.