Search

Abstract Multimodal deep learning plays a pivotal role in supporting the processing and learning of diverse data types within the realm of artificial intelligence generated content (AIGC). However, most photonic neuromorphic processors for deep learning can only handle a single data modality (either vision or audio) due to the lack of abundant parameter training in optical domain. Here, we propose and demonstrate a trainable diffractive optical neural network (TDONN) chip based on on-chip diffractive optics with massive tunable elements to address these constraints. The TDONN chip includes one input layer, five hidden layers, and one output layer, and only one forward propagation is required to obtain the inference results without frequent optical-electrical conversion. The customized stochastic gradient descent algorithm and the drop-out mechanism are developed for photonic neurons to realize in situ training and fast convergence in the optical domain. The TDONN chip achieves a potential throughput of 217.6 tera-operations per second (TOPS) with high computing density (447.7 TOPS/mm2), high system-level energy efficiency (7.28 TOPS/W), and low optical latency (30.2 ps). The TDONN chip has successfully implemented four-class classification in different modalities (vision, audio, and touch) and achieve 85.7% accuracy on multimodal test sets. Our work opens up a new avenue for multimodal deep learning with integrated photonic processors, providing a potential solution for low-power AI large models using photonic technology.

Mycobacterium tuberculosis (Mtb) is an intracellular bacterium that causes a highly contagious and potentially lethal tuberculosis (TB) in humans. It can maintain a dormant TB infection within the host. DosR (dormancy survival regulator) (Rv3133c) has been recognized as one of the key transcriptional proteins regulating bacterial dormancy and participating in various metabolic processes. In this study, we extensively investigate the still not well-comprehended role and mechanism of DosR in Mycobacterium bovis (M. bovis) Bacillus Calmette-Guérin (BCG) through a combined omics analysis. Our study finds that deleting DosR significantly affects the transcriptional levels of 104 genes and 179 proteins. Targeted metabolomics data for amino acids indicate that DosR knockout significantly upregulates L-Aspartic acid and serine synthesis, while downregulating seven other amino acids, including L-histidine and lysine. This suggests that DosR regulates amino acid synthesis and metabolism. Taken together, these findings provide molecular and metabolic bases for DosR effects, suggesting that DosR may be a novel regulatory target.

The Peiligang culture (ca. 9000–7000 cal. BP) represents the first Neolithic settlements in the middle Yellow River Valley, marking the beginning of millet and rice farming in the region. While previous studies have focused primarily on identifying cultivated cereals, less attention has been given to plant harvesting and processing practices or environmental conditions. To address this gap, we present new phytolith data from the Peiligang site (8000–7600 cal. BP) to make three key contributions. First, we show that the Peiligang people utilized two microhabitats: hillslopes for dryland millet cultivation and alluvial plain for wetland resources. Second, we combine our findings with other archaeological evidence to reconstruct the environmental conditions of the Peiligang site, suggesting that it was a water-rich habitat. Finally, by analyzing phytolith remains of plant processing waste in middens, we reconstruct how people harvested and processed millets and rice at the site. This study sheds light on the plant-based subsistence strategies employed by the Peiligang people and offers insights into the environmental factors that contributed to the development of early farming in the middle Yellow River Valley.

Vitamins are essential nutrients and key cofactors of enzymes that regulate cellular metabolism, and also activate the immune system. Recent studies have shown that vitamin B1 (VB1) and vitamin C (Vc) can inhibit Mycobacterium tuberculosis growth, but the precise mechanism is still not well understood. In the present study, we have used RNA-sequencing (RNA-seq), liquid chromatography coupled to mass spectrometry (LC-MS) and single-molecule real-time (SMRT) sequencing to analyze the transcriptional, metabolic and methylation profiles of Mycobacterium bovis BCG when treated with VB1 and Vc. Our results show that, after vitamin treatment, variant metabolites were mainly clustered in pathways related to amino acid metabolism. Treatment with both vitamins significantly up-regulated the gene encoding cysteine synthase A. Additionally, only BCG that was treated with VC showed m4c modifications. Genes harboring this methylation were up-regulated, suggesting that m4c methylation can promote gene transcription to some extent. Overall, this study contributes to the understanding of the effects of VB1 and VC, and suggests that these vitamins constitute potential anti-tuberculosis drugs.