Ismail Sayin, Jacqueline Oien, Deepjyoti Ghosh
et al.
Loss of autoreactive B cell tolerance and production of autoantibodies have been observed in patients undergoing chronic rejection. However, how autoreactive B cells lose their tolerance, and the role of autoreactive antibodies in chronic rejection remains an enigma. Recent observations from our laboratory that autoreactive B cells with a transcriptome profile of innate B cells accumulate in transplanted kidneys diagnosed with antibody-mediated rejection led us to hypothesize that (i) the loss of autoreactive B cell tolerance occurs primarily in the inflamed kidney and (ii) autoantibodies are contributing to graft rejection. In this study, we developed acute and chronic mouse kidney transplant models to test these hypotheses. Donor-specific antibody responses were assayed using donor MHC coated multiplex beads. Autoantibodies were detected with Hep-2 autoantibody detection kit and quantified with an image analysis pipeline in CellProfiler. The presence of autoantibody secreting cells was confirmed by graft tissue culture, and accumulation of autoreactive B cells was confirmed by ex vivo Nojima cultures. We found that autoantibodies were produced during chronic and acute kidney allograft rejection, even when donor MHC-specific antibody was inhibited with CTLA-4Ig. Remarkably, autoantibodies specific for cytoplasmic, nuclear and nucleolar antigens dynamically changed over time, suggesting rapid epitope spreading. Furthermore, there was an enrichment of autoreactive B cells and autoantibody secreting cells in the graft. Taken together, our mouse models complement data from rejecting biopsies, demonstrating a loss of autoreactive B cell tolerance and accumulation of autoantibodies within rejecting allografts. Supported by grant from NIH R01AI148705
Future colliders are an essential component of a strategic vision for particle physics. Conceptual studies and technical developments for several exciting future collider options are underway internationally. In order to realize a future collider, a concerted accelerator R\&D program is required. The U.S. HEP accelerator R\&D program currently has no direct effort in collider-specific R\&D area. This shortcoming greatly compromises the U.S. leadership role in accelerator and particle physics. In this white paper, we propose a new national accelerator R\&D program on future colliders and outline the important characteristics of such a program.
Deep Learning approaches are becoming the go-to methods for data analysis in High Energy Physics (HEP). Nonetheless, most physics-inspired modern architectures are computationally inefficient and lack interpretability. This is especially the case with jet tagging algorithms, where computational efficiency is crucial considering the large amounts of data produced by modern particle detectors. In this work, we present a novel, versatile and transparent framework for jet representation; invariant to Lorentz group boosts, which achieves high accuracy on jet tagging benchmarks while being orders of magnitudes faster to train and evaluate than other modern approaches for both supervised and unsupervised schemes.
Analysis cyberinfrastructure refers to the combination of software and computer hardware used to support late-stage data analysis in High Energy Physics (HEP). For the purposes of this white paper, late-stage data analysis refers specifically to the step of transforming the most reduced common data format produced by a given experimental collaboration (for example, nanoAOD for the CMS experiment) into histograms. In this white paper, we reflect on observations gathered from a recent experience with data analysis using a recent, python-based analysis framework, and extrapolate these experiences though the High-Luminosity LHC era as way of highlighting potential R\&D topics in analysis cyberinfrastructure.
Students from under-represented populations, including those at minority serving institutions have traditionally faced many barriers that have resulted in their being under-represented in High Energy Physics. These barriers include lack of research infrastructure and opportunities, insufficient mentoring, lack of support networks, and financial hardship, among many others. Recently the U.S. CMS Collaboration launched a pilot program U.S. CMS - PURSUE (Program for Undergraduate Research SUmmer Experience) to address these barriers. A 10-week paid internship program, the very first of its kind in an HEP experiment, was organised during the summer of 2022. Students were selected predominantly from Minority Serving Institutions with no research program in HEP. This pilot program provided a structured hands-on research experience under the mentor-ship of U.S. CMS scientists from several collaborating institutions. In addition to emphasis on hands-on research, the program offered a set of software training modules for the first few weeks. These were interleaved with a series of lectures every week covering a broad range of topics. The students were exposed to cutting-edge particle physics research and developed a broad set of skills in software, computing, data science, and machine learning. The modality of this program was virtual, due to the unknown circumstances following the pandemic. There is plan to continue the internship program annually, with in-person training and research participation. In this paper, we describe the experience with the pilot program U.S. CMS - PURSUE.
There have been a number of recent proposals to enhance the performance of machine learning strategies for collider physics by combining many distinct events into a single ensemble feature. To evaluate the efficacy of these proposals, we study the connection between single-event classifiers and multi-event classifiers under the assumption that collider events are independent and identically distributed (IID). We show how one can build optimal multi-event classifiers from single-event classifiers, and we also show how to construct multi-event classifiers such that they produce optimal single-event classifiers. This is illustrated for a Gaussian example as well as for classification tasks relevant for searches and measurements at the Large Hadron Collider. We extend our discussion to regression tasks by showing how they can be phrased in terms of parametrized classifiers. Empirically, we find that training a single-event (per-instance) classifier is more effective than training a multi-event (per-ensemble) classifier, as least for the cases we studied, and we relate this fact to properties of the loss function gradient in the two cases. While we did not identify a clear benefit from using multi-event classifiers in the collider context, we speculate on the potential value of these methods in cases involving only approximate independence, as relevant for jet substructure studies.