Torrey Pines Institute for Molecular Studies

About: Torrey Pines Institute for Molecular Studies is a nonprofit organization based out in San Diego, California, United States. It is known for research contribution in the topics: Antigen & T cell. The organization has 2323 authors who have published 2217 publications receiving 112618 citations.

Extending Adversarial Attacks and Defenses to Deep 3D Point Cloud Classifiers

Abstract: 3D object classification and segmentation using deep neural networks has been extremely successful. As the problem of identifying 3D objects has many safety-critical applications, the neural networks have to be robust against adversarial changes to the input data set. There is a growing body of research on generating human-imperceptible adversarial attacks and defenses against them in the 2D image classification domain. However, 3D objects have various differences with 2D images, and this specific domain has not been rigorously studied so far. We present a preliminary evaluation of adversarial attacks on deep 3D point cloud classifiers, namely PointNet and PointNet++, by evaluating both white-box and black-box adversarial attacks that were proposed for 2D images and extending those attacks to reduce the perceptibility of the perturbations in 3D space. We also show the high effectiveness of simple defenses against those attacks by proposing new defenses that exploit the unique structure of 3D point clouds. Finally, we attempt to explain the effectiveness of the defenses through the intrinsic structures of both the point clouds and the neural network architectures. Overall, we find that networks that process 3D point cloud data are weak to adversarial attacks, but they are also more easily defensible compared to 2D image classifiers. Our investigation will provide the groundwork for future studies on improving the robustness of deep neural networks that handle 3D data.

Immunogenicity. I. Use of Peptide Libraries to Identify Epitopes That Activate Clonotypic CD4+ T Cells and Induce T Cell Responses to Native Peptide Ligands

Abstract: Recent studies have demonstrated the utility of synthetic combinatorial libraries for the rapid identification of peptide ligands that stimulate clonotypic populations of T cells. Here we screen a decapeptide combinatorial library arranged in a positional scanning format with two different clonotypic populations of CD4+ T cells to identify peptide epitopes that stimulate proliferative responses by these T cells in vitro. An extensive collection of mimic peptide sequences was synthesized and used to explore the fine specificity of TCR/peptide/MHC interactions. We also demonstrate that many of these deduced ligands are not only effective immunogens in vivo, but are capable of inducing T cell responses to the original native ligands used to generate the clones. These results have significant implications for considerations of T cell specificity and the design of peptide vaccines for infectious disease and cancer using clinically relevant T cell clones of unknown specificity.

Specific control of immunity by regulatory CD8 T cells.

Abstract: T lymphocytes with dedicated suppressor function (Treg) play a crucial role in the homeostatic control of immunity in the periphery. Several Treg phenotypes have now been identified in the CD4 and CD8 T cell populations, suggesting their down-regulatory function in both human and animal models of autoimmunity, transplantation and tumor immunity. Here we will focus on the CD8 Treg population and their ability to specifically inhibit a pathogenic autoimmune response. This review will detail the current advances in the knowledge of CD8 Treg in the context of antigen specificity, phenotype, MHC restriction, mechanism of action, and priming.

Extending Adversarial Attacks and Defenses to Deep 3D Point Cloud Classifiers

Abstract: 3D object classification using deep neural networks has been extremely successful. As the problem of identifying 3D objects has many safety-critical applications, the neural networks have to be robust against adversarial changes to the input data set. We present a preliminary evaluation of adversarial attacks on 3D point cloud classifiers by evaluating adversarial attacks that were proposed for 2D images, and extending those attacks to reduce the perceptibility of the perturbations in 3D space. We also show the effectiveness of simple defenses against those attacks. Finally, we attempt to explain the effectiveness of the defenses through the intrinsic structures of both the point clouds and the neural networks. Overall, we find that 3D point cloud classifiers are weak to adversarial attacks, but they are also more easily defensible compared to 2D image classifiers. Our investigation will provide the groundwork for future studies on improving the robustness of deep neural networks that handle 3D data.

Screening of Differentially Amplified cDNA Products from RNA Arbitrarily Primed PCR Fingerprints Using Single Strand Conformation Polymorphism (SSCP) Gels

Abstract: Arbitrarily primed PCR fingerprinting of RNA and differential display resolved on an acrylamide gel has been extensively used to detect differentially expressed RNAs. However, after a differentially amplified product is detected the next steps are labor-intensive: a small portion of the fingerprinting gel that contains the differentially amplified product is cut out, reamplified and the correct product is determined, typically by cloning and sequencing what is often a mixture of products of similar size. Here we use a native acrylamide gel to separate DNAs in the reamplified mixture based on single-stranded conformation polymorphisms. Reamplifications are performed for the region carrying the differentially amplified product and a corresponding region from an adjacent lane where the product is less prominent or not visible. Denaturation of the reamplified DNA followed by side-by-side comparison on an SSCP gel allows the classification of reamplified material into (i) those that can be directly cloned because the differentially amplified product is relatively pure, (ii) those that need to be reamplified from the SSCP gel before cloning and (iii) those that are too complex for further study. This screen should save considerable effort now wasted on directly cloning unsuitable products from RNA fingerprinting experiments. An example is presented of cloning a gene differentially expressed in Trypanosoma brucei life cycle.