Browse Articles
Discover research articles across all indexed journals
Rapid learning with phase-change memory-based in-memory computing through learning-to-learn
Abstract There is a growing demand for low-power, autonomously learning artificial intelligence (AI) systems that can be applied at the edge and rapidly adapt to the specific situation at deployment site. However, current AI models struggle in such scenarios, often requiring extensive fine-tuning, computational resources, and data. In contrast, humans can effortlessly adjust to new tasks by transferring knowledge from related ones. The concept of learning-to-learn (L2L) mimics this process and enables AI models to rapidly adapt with only little computational effort and data. In-memory computing neuromorphic hardware (NMHW) is inspired by the brain’s operating principles and mimics its physical co-location of memory and compute. In this work, we pair L2L with in-memory computing NMHW based on phase-change memory devices to build efficient AI models that can rapidly adapt to new tasks. We demonstrate the versatility of our approach in two scenarios: a convolutional neural network performing image classification and a biologically-inspired spiking neural network generating motor commands for a real robotic arm. Both models rapidly learn with few parameter updates. Deployed on the NMHW, they perform on-par with their software equivalents. Moreover, meta-training of these models can be performed in software with high-precision, alleviating the need for accurate hardware models.
Spatial integration of multi-omics single-cell data with SIMO
An atlas of metabolites driving chemotaxis in prokaryotes
Copy number amplification of FLAD1 promotes the progression of triple-negative breast cancer through lipid metabolism
Vaccine-induced T cell receptor T cell therapy targeting a glioblastoma stemness antigen
Abstract T cell receptor-engineered T cells (TCR-T) could be advantageous in glioblastoma by allowing safe and ubiquitous targeting of the glioblastoma-derived peptidome. Protein tyrosine phosphatase receptor type Z1 (PTPRZ1), is a clinically targetable glioblastoma antigen associated with glioblastoma cell stemness. Here, we identify a therapeutic HLA-A*02-restricted PTPRZ1-reactive TCR retrieved from a vaccinated glioblastoma patient. Single-cell sequencing of primary brain tumors shows PTPRZ1 overexpression in malignant cells, especially in glioblastoma stem cells (GSCs) and astrocyte-like cells. The validated vaccine-induced TCR recognizes the endogenously processed antigen without off-target cross-reactivity. PTPRZ1-specific TCR-T (PTPRZ1-TCR-T) kill target cells antigen-specifically, and in murine experimental brain tumors, their combined intravenous and intracerebroventricular administration is efficacious. PTPRZ1-TCR-T maintain stem cell memory phenotype in vitro and in vivo and lyse all examined HLA-A*02 + primary glioblastoma cell lines with a preference for GSCs and astrocyte-like cells. In summary, we demonstrate the proof of principle to employ TCR-T to treat glioblastoma.
Nutrient status alters developmental fates via a switch in mitochondrial homeodynamics
Sustainable synthesis of α-ketoglutaric and methanetriacetic acids from biomass feedstocks
Transcriptome size matters for single-cell RNA-seq normalization and bulk deconvolution
Abstract The variation of transcriptome size across cell types significantly impacts single-cell RNA sequencing (scRNA-seq) data normalization and bulk RNA-seq cellular deconvolution, yet this intrinsic feature is often overlooked. Here we introduce ReDeconv, a computational algorithm that incorporates transcriptome size into scRNA-seq normalization and bulk deconvolution. ReDeconv introduces a scRNA-seq normalization approach, Count based on Linearized Transcriptome Size (CLTS), which corrects differential expressed genes typically misidentified by standard count per 10 K normalization, as confirmed by orthogonal validations. By maintaining transcriptome size variation, CLTS-normalized scRNA-seq enhances the accuracy of bulk deconvolution. Additionally, ReDeconv mitigates gene length effects and models expression variances, thereby improving deconvolution outcomes, particularly for rare cell types. Evaluated with both synthetic and real datasets, ReDeconv surpasses existing methods in precision. ReDeconv alters the practice and provides a new standard for scRNA-seq analyses and bulk deconvolution. The software packages and a user-friendly web portal are available.
Proteomic profiling identifies muscle-invasive bladder cancers with distinct biology and responses to platinum-based chemotherapy
The role of viral interaction in household transmission of symptomatic influenza and respiratory syncytial virus
Prompt injection attacks on vision language models in oncology
Abstract Vision-language artificial intelligence models (VLMs) possess medical knowledge and can be employed in healthcare in numerous ways, including as image interpreters, virtual scribes, and general decision support systems. However, here, we demonstrate that current VLMs applied to medical tasks exhibit a fundamental security flaw: they can be compromised by prompt injection attacks. These can be used to output harmful information just by interacting with the VLM, without any access to its parameters. We perform a quantitative study to evaluate the vulnerabilities to these attacks in four state of the art VLMs: Claude-3 Opus, Claude-3.5 Sonnet, Reka Core, and GPT-4o. Using a set of N = 594 attacks, we show that all of these models are susceptible. Specifically, we show that embedding sub-visual prompts in manifold medical imaging data can cause the model to provide harmful output, and that these prompts are non-obvious to human observers. Thus, our study demonstrates a key vulnerability in medical VLMs which should be mitigated before widespread clinical adoption.
Electronic descriptors for designing high-entropy alloy electrocatalysts by leveraging local chemical environments
PPARα-mediated lipid metabolism reprogramming supports anti-EGFR therapy resistance in head and neck squamous cell carcinoma
The ubiquitin ligase Pellino1 targets STAT3 to regulate macrophage-mediated inflammation and tumor development
Abstract Receptor-mediated signaling could be modulated by ubiquitination of pathway intermediates, but the role of such modification in the pathogenesis of inflammation and inflammation-related cancer is lesser known. The ubiquitin ligase Pellino1 has been shown to modulate immune signals by enabling various immune cells to respond to their receptor signals effectively. Here, we show that Pellino1 levels are elevated in patients with colitis, patients with colitis-associated colon cancer (CAC), and murine models of these conditions. In a monocyte-specific Pellino1 knock-out mouse model, we find reduced macrophage migration and activation, leading to attenuated development of colitis and CAC in male mice. Mechanistically, Pellino1 targets STAT3 for lysine 63-mediated ubiquitination, resulting in pathogenic activation of STAT3 signaling. Taken together, our findings reveal a macrophage-specific ubiquitination signaling axis in colitis and CAC development and suggest that Pellino1 is a potential candidate for treating chronic inflammation and inflammation-related cancer.
Analysis of eIF4E-family members in fungi contributes to their classification in eukaryotes
miR-142 deficit in T cells during blast crisis promotes chronic myeloid leukemia immune escape
Abstract We reported that an acquired miR-142 deficit transforms chronic phase (CP) chronic myeloid leukemia (CML) leukemic stem cells (LSCs) into blast crisis (BC) LSCs. Given the role of miR-142 in the development and activity of the immune system, we postulated that this deficit also promotes LSC immune escape. Herein, we report on IL-6-driven miR-142 deficit occurring in T cells during BC transformation. In CML murine models, miR-142 deficit impairs thymic differentiation of lymphoid-primed multipotent progenitors (LMPP) into T cells and prevents T cells’ metabolic reprogramming, thereby leading to loss of T cells and leukemia immune escape. Correcting miR-142 deficit with a miR-142 mimic compound (M-miR-142), alone or in combination with immune checkpoint antibodies, restores T cell number and immune activity, leading to LSC elimination and prolonged survival of BC CML murine and patient-derived xenograft models. These observations may open new therapeutic opportunities for BC CML and other myeloid malignancies.