ProteinMAE: masked autoencoder for protein surface self-supervised learning.

SUMMARY: The biological functions of proteins are determined by the chemical and geometric properties of their surfaces. Recently, with the booming progress of deep learning, a series of learning-based surface descriptors have been proposed and achieved inspirational performance in many tasks such as protein design, protein-protein interaction prediction, etc. However, they are still limited by the problem of label scarcity, since the labels are typically obtained through wet experiments. Inspired by the great success of self-supervised learning in natural language processing and computer vision, we introduce ProteinMAE, a self-supervised framework specifically designed for protein surface representation to mitigate label scarcity. Specifically, we propose an efficient network and utilize a large number of accessible unlabeled protein data to pretrain it by self-supervised learning. Then we use the pretrained weights as initialization and fine-tune the network on downstream tasks. To demonstrate the effectiveness of our method, we conduct experiments on three different downstream tasks including binding site identification in protein surface, ligand-binding protein pocket classification, and protein-protein interaction prediction. The extensive experiments show that our method not only successfully improves the network's performance on all downstream tasks, but also achieves competitive performance with state-of-the-art methods. Moreover, our proposed network also exhibits significant advantages in terms of computational cost, which only requires less than a tenth of memory cost of previous methods. AVAILABILITY AND IMPLEMENTATION: https://github.com/phdymz/ProteinMAE.

Robust Point Cloud Registration Framework Based on Deep Graph Matching.

3D point cloud registration is a fundamental problem in computer vision and robotics. Recently, learning-based point cloud registration methods have made great progress. However, these methods are sensitive to outliers, which lead to more incorrect correspondences. In this paper, we propose a novel deep graph matching-based framework for point cloud registration. Specifically, we first transform point clouds into graphs and extract deep features for each point. Then, we develop a module based on deep graph matching to calculate a soft correspondence matrix. By using graph matching, not only the local geometry of each point but also its structure and topology in a larger range are considered in establishing correspondences, so that more correct correspondences are found. We train the network with a loss directly defined on the correspondences, and in the test stage the soft correspondences are transformed into hard one-to-one correspondences so that registration can be performed by a correspondence-based solver. Furthermore, we introduce a transformer-based method to generate edges for graph construction, which further improves the quality of the correspondences. Extensive experiments on object-level and scene-level benchmark datasets show that the proposed method achieves state-of-the-art performance.

Global optimization point-set registration based on translation/rotation decoupling for image-guided surgery applications.

PURPOSE: In image-guided surgery systems, image-to-patient spatial registration is to get the spatial transformation between the image space and the actual operating space. Although the image-to-patient spatial registration methods using paired point or surface matching are used in some image-guided neurosurgery systems, the key problem is that the global optimization registration result cannot be achieved. Therefore, this paper proposes a new rotation invariant feature for decoupling rotation and translation space, based on which global optimization point set registration method is proposed. METHODS: The new rotation invariant features, constructed based on the edges and the angles, are the rotation invariant, which has high feature resolution. Some of them are not only the rotation invariant, but also the translation invariant. To obtain the global optimal solution, branch and bound search strategy is used to search the parameter space of the translation and the computational cost is reduced simultaneously. The registration accuracy of the spatial registration method is analyzed by comparing the difference between the estimated transform and the standard transform to calculate the registration error. RESULTS: To validate the performance of the spatial registration method proposed, the registration performance was analyzed by comparing the experimental results with the results of the two mainstream registration methods (the iterative closest point [ICP] registration method and the coherent point drift method). In the experiments, the comparison was based on the registration accuracy and the execution time. We show our registration method can obtain higher accuracy in a shorter time in most cases. At the same time, when using ICP to further refine our results, the ICP method can converge in a very short time, which also shows that our method provides a good initial pose for the ICP method and can help the ICP converge to the global optimal solution faster. Our method can achieve an average rotation error of 0.124 degrees and an average translation error of 0.38 mm on 10 clinical data. CONCLUSIONS: The results reveal that the surface registration method based on translation rotation decoupling can achieve superior performance regarding both the registration accuracy and the time efficiency in the image-to-patient spatial registration.