Convolutional neural networks (CNNs) have exhibited state-of-the-art performance in various audio classification tasks. However, their real-time deployment remains a challenge on resource-constrained devices such as embedded systems. In this paper, we analyze how the performance of large-scale pre-trained audio neural networks designed for audio pattern recognition changes when deployed on a hardware such as a Raspberry Pi. We empirically study the role of CPU temperature, microphone quality and audio signal volume on performance. Our experiments reveal that the continuous CPU usage results in an increased temperature that can trigger an automated slowdown mechanism in the Raspberry Pi, impacting inference latency. The quality of a microphone, specifically with affordable devices such as the Google AIY Voice Kit, and audio signal volume, all affect the system performance. In the course of our investigation, we encounter substantial complications linked to library compatibility and the unique processor architecture requirements of the Raspberry Pi, making the process less straightforward compared to conventional computers (PCs). Our observations, while presenting challenges, pave the way for future researchers to develop more compact machine learning models, design heat-dissipative hardware, and select appropriate microphones when AI models are deployed for real-time applications on edge devices.
Convolutional neural networks (CNNs) are commonplace in
high-performing solutions to many real-world problems, such as
audio classification. CNNs have many parameters and filters,
with some having a larger impact on the performance than others.
This means that networks may contain many unnecessary filters,
increasing a CNN’s computation and memory requirements while
providing limited performance benefits. To make CNNs more
efficient, we propose a pruning framework that eliminates filters
with the highest “commonality”. We measure this commonality
using the graph-theoretic concept of centrality. We hypothesise
that a filter with a high centrality should be eliminated as it
represents commonality and can be replaced by other filters without
affecting the performance of a network much. An experimental
evaluation of the proposed framework is performed on acoustic
scene classification and audio tagging. On the DCASE 2021 Task
1A baseline network, our proposed method reduces computations
per inference by 71% with 50% fewer parameters at less than a
two percentage point drop in accuracy compared to the original
network. For large-scale CNNs such as PANNs designed for audio
tagging, our method reduces 24% computations per inference with
41% fewer parameters at a slight improvement in performance.
Cells secrete extracellular vesicles (EVs) through various biogenesis pathways, resulting in distinct molecular compositions even when originating from the same cells. Analyzing individual EVs is challenging due to the need to overcome issues such as nanoscale size, heterogeneity of EVs, and measurement accuracy. Overcoming these challenges in EV research not only advances the field of biopsy but also enables progress in the biological research of EVs. In this study, we analyze quantifying membrane proteins using a novel methodology based on single-molecule fluorescence spectroscopy. We used TIRF based fluorescence imaging or Fluorescence Correlation Spectroscopy (FCS) to observe antibodies or aptamers binding to single EVs. Our results demonstrate the quantification of various membrane proteins, including CD63 and CD81. Consequently, we develop a membrane protein quantification assay for individual EVs using single-molecule and particle fluorescence imaging spectroscopy. Through this single molecule assay, we will elucidate the distribution of membrane proteins in EVs derived from various tumor cells, potentially enabling their use in future biopsies or quantification studies.
Antibody-based therapies offer significant promise for the treatment of various diseases, including cancer and infectious agents. In cancer immunotherapy, these therapies aim to activate the immune system against tumor cells or inhibit their growth. However, the complex nature of cancer necessitates the exploration of multi-target therapeutic approaches, such as bispecific antibodies. These antibodies offer the potential to target multiple cancer types and variants, potentially overcoming resistance mechanisms. Single-domain antibodies (sdAbs) possess unique characteristics, including a compact size, high stability, and specific antigen recognition, making them ideal building blocks for bispecific antibody development. In this study, we describe the development of a readily manageable and cost-effective sdAb discovery platform utilizing rabbit immunization. Briefly, rabbits were immunized with a recombinant CD3 epsilon domain to elicit an immune response. Subsequently, antigen-specific sdAbs were isolated from the immunoglobulin heavy chain variable region. This accessible sdAb discovery platform has the potential to be applied in a variety of fields, including biological research and therapeutic development.
The development of bispecific antibodies that redirect the cytotoxic activity of CD3+ T cells against tumors is a promising immunotherapy strategy for hematological malignancies and solid tumors. The discovery and development of novel anti-CD3 antibodies are key to the efficacy and safety of the bispecific antibodies. Camelid-derived nanobodies have significant potential in bispecific antibodies as T cell engagers due to their small size, low production cost, high stability, and antigen specificity.
In this study, we identified variable nanobodies specific for the CD3 by immunizing a Vicugna pacos (alpaca) with recombinant human CD3 epsilon domain. In particular, the anti-CD3 Nb 1D10 clone showed a moderate affinity with recombinant human and Macaca fascicularis (cynomolgus monkey) CD3 and binding to CD3+ cell lines. These results highlight the potential of the nanobody for the development of nanobody-based bispecific antibodies as a T cell engager.
Cannabinoid receptor 1 (CB1R), a G protein-coupled receptor, plays a critical role in regulating appetite and exhibits increased expression in peripheral insulin-target tissues during obesity. This suggests its potential involvement in obesity-induced pro-inflammatory responses. Selective targeting of peripheral CB1R could offer a novel therapeutic approach to break the link between insulin resistance and metabolic inflammation. However, The widespread distribution of CB1R, including the central nervous system (CNS), presents a challenge. CNS-directed CB1R blockade can lead to severe psychological effects like depression and suicidality. This study investigates the development of peripherally-restricted, high-affinity single-domain antibodies (sdAbs) targeting CB1R for selective appetite modulation. We employed an APG-solubilized, recombinant CB1R for rabbit immunization. Antigen-specific sdAbs were subsequently isolated from the immunoglobulin heavy chain variable region. Biopanning of the resulting phage display library was conducted to identify sdAbs with high binding affinity for CB1R. Our findings demonstrate the development of CB1R-specific sdAbs, potentially offering a novel and targeted strategy for obesity management with minimized CNS exposure and reduced risk of associated psychiatric side effects
Bispecific antibodies (BsAbs) that redirect cytotoxic T lymphocytes (CTLs) via CD3 engagement represent a promising approach in cancer immunotherapy. Here, we explore the development of VNAR-based BsAbs for improved tumor targeting. Shark-derived VNARs possess unique properties such as small size, high stability, and antigen specificity, making them ideal candidates for T cell engagement. We immunized banded houndsharks (Triakis scyllium) with the recombinant human CD3 epsilon domain to isolate VNARs that specifically recognize CD3. The isolated anti-CD3 VNARs showed moderate affinity for recombinant human CD3 and bound to CD3+ cell lines. These results suggest the potential of these VNARs for the construction of VNAR-based BsAbs as T cell engagers, providing a novel avenue for cancer immunotherapy.
The neonatal Fc receptor (FcRn) extends the serum half-life of immunoglobulins G (IgG) through a pH-dependent interaction that protects and recycles IgG during intracellular trafficking. To understand the molecular basis of the prolonged half-life observed with the Fc variant KU-1, this study employs X-ray crystallography and in silico modeling to elucidate the mode of interaction (MOI) between FcRn and KU-1. Deciphering this interaction has the potential to significantly impact the development of KU-1-based biopharmaceuticals, paving the way for enhanced therapeutic efficacy and safety profiles.
CD27, a costimulatory molecule of the TNF receptor superfamily expressed on T lymphocytes, plays a critical role in regulating T cell survival, differentiation, and effector function. Upon binding its ligand, CD70, CD27 signaling enhances T cell proliferation and differentiation into effector and memory T cells. This agonistic activity positions CD27 as a promising target for immunomodulatory cancer therapy. This study investigates the development of therapeutic agents targeting the CD27 for cancer immunotherapy. We employed a recombinant human CD27 extracellular domain to immunize an alpaca (Vicugna pacos), generating antigen-specific single-domain antibodies (nanobodies). Through biopanning using the immunized phage display library, we identified the anti-CD27 cpNb4 clone exhibiting high binding affinity for recombinant human CD27 and specific binding to cell-surface CD27 expressed on CHO-K1 cells. These findings establish cpNb4 as a promising candidate for further investigation, potentially paving the way for its integration into combination immunomodulatory cancer therapy regimens.
Cellular senescence is a permanent cell proliferation arrest compromising cell regeneration and tissue repair process which gradually leads to age-related disorders. The driving factor is considered to be senescence-associated secretory phenotype(SASP) a diverse pro-inflammatory secretory factors exerted to nearby cells resulting in normal cell aging. Targeted therapies with complex drug that can selectively modulate these cells offer promising avenues for anti-aging interventions. In this study, we developed extracellular vesicles(EVs) with specific ligands as a novel drug carrier system aimed to selectively target senescent cells. By attaching ligands on the surface of EVs, we enhanced their affinity for senescent cells over normal cells and the targeting efficiency was assessed using fluorescence-activated cell sorting(FACs). The results demonstrated a significantly higher uptake of modified EVs in senescent cells which suggest that this selective delivery can effectively serve as a precision drug carrier. Future work will focus on loading therapeutic agents that exhibit senolytic or senomorphic activities to these ligand-modified EVs aiming to selectively attenuate cellular senescence.