Search

Importance: Reinforcement learning (RL) represents a pivotal avenue within natural language processing (NLP), offering a potent mechanism for acquiring optimal strategies in task completion. This literature review studies various NLP applications where RL has demonstrated efficacy, with notable applications in healthcare settings., Objectives: To systematically explore the applications of RL in NLP, focusing on its effectiveness in acquiring optimal strategies, particularly in healthcare settings, and provide a comprehensive understanding of RL's potential in NLP tasks., Materials and Methods: Adhering to the PRISMA guidelines, an exhaustive literature review was conducted to identify instances where RL has exhibited success in NLP applications, encompassing dialogue systems, machine translation, question-answering, text summarization, and information extraction. Our methodological approach involves closely examining the technical aspects of RL methodologies employed in these applications, analyzing algorithms, states, rewards, actions, datasets, and encoder-decoder architectures., Results: The review of 93 papers yields insights into RL algorithms, prevalent techniques, emergent trends, and the fusion of RL methods in NLP healthcare applications. It clarifies the strategic approaches employed, datasets utilized, and the dynamic terrain of RL-NLP systems, thereby offering a roadmap for research and development in RL and machine learning techniques in healthcare. The review also addresses ethical concerns to ensure equity, transparency, and accountability in the evolution and application of RL-based NLP technologies, particularly within sensitive domains such as healthcare., Discussion: The findings underscore the promising role of RL in advancing NLP applications, particularly in healthcare, where its potential to optimize decision-making and enhance patient outcomes is significant. However, the ethical challenges and technical complexities associated with RL demand careful consideration and ongoing research to ensure responsible and effective implementation., Conclusions: By systematically exploring RL's applications in NLP and providing insights into technical analysis, ethical implications, and potential advancements, this review contributes to a deeper understanding of RL's role for language processing., (© The Author(s) 2024. Published by Oxford University Press on behalf of the American Medical Informatics Association. All rights reserved. For permissions, please email: journals.permissions@oup.com.)

Fine-tuning pretrained protein language models (PLMs) has emerged as a prominent strategy for enhancing downstream prediction tasks, often outperforming traditional supervised learning approaches. As a widely applied powerful technique in natural language processing, employing parameter-efficient fine-tuning techniques could potentially enhance the performance of PLMs. However, the direct transfer to life science tasks is nontrivial due to the different training strategies and data forms. To address this gap, we introduce SES-Adapter, a simple, efficient, and scalable adapter method for enhancing the representation learning of PLMs. SES-Adapter incorporates PLM embeddings with structural sequence embeddings to create structure-aware representations. We show that the proposed method is compatible with different PLM architectures and across diverse tasks. Extensive evaluations are conducted on 2 types of folding structures with notable quality differences, 9 state-of-the-art baselines, and 9 benchmark data sets across distinct downstream tasks. Results show that compared to vanilla PLMs, SES-Adapter improves downstream task performance by a maximum of 11% and an average of 3%, with significantly accelerated convergence speed by a maximum of 1034% and an average of 362%, the training efficiency is also improved by approximately 2 times. Moreover, positive optimization is observed even with low-quality predicted structures. The source code for SES-Adapter is available at https://github.com/tyang816/SES-Adapter.

Protein language models (PLMs) play a dominant role in protein representation learning. Most existing PLMs regard proteins as sequences of 20 natural amino acids. The problem with this representation method is that it simply divides the protein sequence into sequences of individual amino acids, ignoring the fact that certain residues often occur together. Therefore, it is inappropriate to view amino acids as isolated tokens. Instead, the PLMs should recognize the frequently occurring combinations of amino acids as a single token. In this study, we use the byte-pair-encoding algorithm and unigram to construct advanced residue vocabularies for protein sequence tokenization, and we have shown that PLMs pre-trained using these advanced vocabularies exhibit superior performance on downstream tasks when compared to those trained with simple vocabularies. Furthermore, we introduce PETA, a comprehensive benchmark for systematically evaluating PLMs. We find that vocabularies comprising 50 and 200 elements achieve optimal performance. Our code, model weights, and datasets are available at https://github.com/ginnm/ProteinPretraining . SCIENTIFIC CONTRIBUTION: This study introduces advanced protein sequence tokenization analysis, leveraging the byte-pair-encoding algorithm and unigram. By recognizing frequently occurring combinations of amino acids as single tokens, our proposed method enhances the performance of PLMs on downstream tasks. Additionally, we present PETA, a new comprehensive benchmark for the systematic evaluation of PLMs, demonstrating that vocabularies of 50 and 200 elements offer optimal performance., (© 2024. The Author(s).)

Accurately modeling the protein fitness landscapes holds great importance for protein engineering. Pre-trained protein language models have achieved state-of-the-art performance in predicting protein fitness without wet-lab experimental data, but their accuracy and interpretability remain limited. On the other hand, traditional supervised deep learning models require abundant labeled training examples for performance improvements, posing a practical barrier. In this work, we introduce FSFP, a training strategy that can effectively optimize protein language models under extreme data scarcity for fitness prediction. By combining meta-transfer learning, learning to rank, and parameter-efficient fine-tuning, FSFP can significantly boost the performance of various protein language models using merely tens of labeled single-site mutants from the target protein. In silico benchmarks across 87 deep mutational scanning datasets demonstrate FSFP's superiority over both unsupervised and supervised baselines. Furthermore, we successfully apply FSFP to engineer the Phi29 DNA polymerase through wet-lab experiments, achieving a 25% increase in the positive rate. These results underscore the potential of our approach in aiding AI-guided protein engineering., (© 2024. The Author(s).)

Generative Large Language Models (LLMs) have achieved significant success in various natural language processing tasks, including Question-Answering (QA) and dialogue systems. However, most models are trained on English data and lack strong generalization in providing answers in Chinese. This limitation is especially evident in specialized domains like traditional Chinese medical QA, where performance suffers due to the absence of fine-tuning and high-quality datasets. To address this, we introduce MedChatZH, a dialogue model optimized for Chinese medical QA based on transformer decoder with LLaMA architecture. Continued pre-training on a curated corpus of Chinese medical books is followed by fine-tuning with a carefully selected medical instruction dataset, resulting in MedChatZH outperforming several Chinese dialogue baselines on a real-world medical dialogue dataset. Our model, code, and dataset are publicly available on GitHub (https://github.com/tyang816/MedChatZH) to encourage further research in traditional Chinese medicine and LLMs., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024 Elsevier Ltd. All rights reserved.)

The COVID-19 pandemic has been accompanied by a proliferation of online misinformation and disinformation about the virus. Combating this 'infodemic' has been identified as one of the top priorities of the World Health Organization, because false and misleading information can lead to a range of negative consequences, including the spread of false remedies, conspiracy theories, and xenophobia. This paper aims to combat the COVID-19 infodemic on multiple fronts, including determining the credibility of information, identifying its potential harm to society, and the necessity of intervention by relevant organizations. We present a prompt-based curriculum learning method to achieve this goal. The proposed method could overcome the challenges of data sparsity and class imbalance issues. Using online social media texts as input, the proposed model can verify content from multiple perspectives by answering a series of questions concerning the text's reliability. Experiments revealed the effectiveness of prompt tuning and curriculum learning in assessing the reliability of COVID-19-related text. The proposed method outperforms typical text classification methods, including fastText and BERT. In addition, the proposed method is robust to the hyperparameter settings, making it more applicable with limited infrastructure resources., Competing Interests: The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (© 2023 Elsevier Ltd. All rights reserved.)

The coronavirus disease (COVID-19) pandemic has already caused enormous damage to the global economy and various industries worldwide, especially the tourism industry. In the post-pandemic era, accurate tourism demand recovery forecasting is a vital requirement for a thriving tourism industry. Therefore, this study mainly focuses on forecasting tourist arrivals from mainland China to Hong Kong. A new direction in tourism demand recovery forecasting employs multi-source heterogeneous data comprising economy-related variables, search query data, and online news data to motivate the tourism destination forecasting system. The experimental results confirm that incorporating multi-source heterogeneous data can substantially strengthen the forecasting accuracy. Specifically, mixed data sampling (MIDAS) models with different data frequencies outperformed the benchmark models., Competing Interests: The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (© 2023 Elsevier Ltd. All rights reserved.)

Introduction: Nighttime crashes account for 74% of pedestrian fatalities in the United States, and reduced visibility is a significant cause of nighttime pedestrian crashes. Maintaining sufficient and uniform roadway lighting is an effective countermeasure to improve pedestrian visibility and prevent nighttime pedestrian crashes and injuries. Previous studies have not quantified the safety effects of roadway photometric patterns (i.e., average lighting level and uniformity) on nighttime pedestrian crashes on roadway segments., Method: This study investigated the association between two roadway photometric criteria (horizontal illuminance mean representing average lighting level and horizontal illuminance standard deviation representing lighting uniformity) and nighttime pedestrian crash occurrence in Florida roadway segments. The matched case-control method was used to decouple the confounding effects between the illuminance mean and standard deviation. Statistically-significant crash modification factors (CMFs) were developed to quantify the safety effects of the mean and standard deviation of horizontal illuminance on nighttime pedestrian crashes., Results: The results show that if the average lighting level on a roadway segment is increased from a low illuminance mean (<0.2 foot-candle [fc]) to a medium illuminance mean [0.2 fc, 0.5 fc], a medium-high illuminance mean (0.5 fc, 1.0 fc], and a high illuminance mean (>1.0 fc), the relative likelihood of nighttime pedestrian crashes on midblock segments in Florida tends to be reduced by 77.5% (CMF = 0.225), 81.2% (CMF = 0.188), and 85.5% (CMF = 0.145), respectively., Practical Applications: A poor uniformity (illuminance standard deviation ≥ 0.52 fc) is likely to increase the relative likelihood of nighttime pedestrian crashes on midblock segments in Florida by 80.3% (CMF = 1.803) compared to good uniformity (illuminance standard deviation < 0.52 fc)., Competing Interests: Declaration of Competing Interest The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: [Zhenyu Wang reports financial support was provided by The Center for Transportation, Equity, Decisions and Dollars (CTEDD) at the University of Texas at Arlington.]., (Copyright © 2023 National Safety Council and Elsevier Ltd. All rights reserved.)

Transition metal dichalcogenides (TMDs) have attracted great attention in valleytronics. Owing to the giant valley coherence at room temperature, valley pseudospin of TMDs open a new degree of freedom to encode and process binary information. The valley pseudospin only exists in non-centrosymmetric TMDs (e.g., monolayer or 3R-stacked multilayer), which is prohibited in conventional centrosymmetric 2H-stacked crystals. Here, we propose a general recipe to generate valley-dependent vortex beams by using a mix-dimensional TMD metasurface composed of nanostructured 2H-stacked TMD crystals and monolayer TMDs. Such an ultrathin TMD metasurface involves a momentum-space polarization vortex around bound states in the continuum (BICs), which can simultaneously achieve strong coupling (i.e., form exciton polaritons) and valley-locked vortex emission. Moreover, we report that a full 3R-stacked TMD metasurface can also reveal the strong-coupling regime with an anti-crossing pattern and a Rabi splitting of 95 meV. The Rabi splitting can be precisely controlled by geometrically shaping the TMD metasurface. Our results provide an ultra-compact TMD platform for controlling and structuring valley exciton polariton, in which the valley information is linked with the topological charge of vortex emission, which may advance valleytronic, polaritonic, and optoelectronic applications.

Deep learning has been widely used for protein engineering. However, it is limited by the lack of sufficient experimental data to train an accurate model for predicting the functional fitness of high-order mutants. Here, we develop SESNet, a supervised deep-learning model to predict the fitness for protein mutants by leveraging both sequence and structure information, and exploiting attention mechanism. Our model integrates local evolutionary context from homologous sequences, the global evolutionary context encoding rich semantic from the universal protein sequence space and the structure information accounting for the microenvironment around each residue in a protein. We show that SESNet outperforms state-of-the-art models for predicting the sequence-function relationship on 26 deep mutational scanning datasets. More importantly, we propose a data augmentation strategy by leveraging the data from unsupervised models to pre-train our model. After that, our model can achieve strikingly high accuracy in prediction of the fitness of protein mutants, especially for the higher order variants (> 4 mutation sites), when finetuned by using only a small number of experimental mutation data (< 50). The strategy proposed is of great practical value as the required experimental effort, i.e., producing a few tens of experimental mutation data on a given protein, is generally affordable by an ordinary biochemical group and can be applied on almost any protein., (© 2023. The Author(s).)

In Yb-Er co-doped upconversion (UC) nanomaterials, upconversion luminescence (UCL) can be modulated to generate multiband UCL emissions by changing the concentration of activator Er 3+ . Nonetheless, the effect of the Er 3+ concentrations on the kinetics of these emissions is still unknown. We here study the single β -NaYF 4 :Yb 3+ /Er 3+ microcrystal (MC) doped with different Er 3+ concentrations by nanosecond time-resolved spectroscopy. Interestingly, different Er 3+ doping concentrations exhibit different UCL emission bands and UCL response rates. At low Er 3+ doping concentrations (1 mol%), multiband emission in β -NaYF 4 :Yb 3+ /Er 3+ (20/1 mol%) MCs could not be observed and the response rate of UCL was slow (5-10 μs) in β -NaYF 4 :Yb 3+ /Er 3+ . Increasing the Er 3+ doping concentration to 10 mol% can shorten the distance between Yb 3+ ions and Er 3+ ions, which promotes the energy transfer between them. β -NaYF 4 :Yb 3+ /Er 3+ (20/10 mol%) can achieve obvious multiband UCL and a quick response rate (0.3 µs). However, a further increase in the Er doping concentration (80 mol%) makes MCs limited by the CR process and cannot achieve the four-photon UC process ( 4 F 5/2 → 2 K 13/2 and 2 H 9/2 → 2 D 5/2 ). Therefore, the result shows that changing the Er 3+ doping concentration could control the energy flow between the different energy levels in Er 3+ , which could affect the response time and UCL emission of the Yb/Er doped rare earth materials. Our work can facilitate the development of fast-response optoelectronics, optical-sensing, and display industries., Competing Interests: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest., (Copyright © 2023 Huang, Zhong, Li, Cui, Yu, Zhao, Xing, Guo and Han.)

Single-domain antibody (sdAb) holds the promising strategies for diverse research and translational applications. Here, we describe a method for the adaptation of the in situ proximity ligation assay ( is PLA) followed by sequencing ( is PLA-seq) to facilitate screening of a high-sensitive, high-throughput sdAb library for a given protein at subcellular and single-cell resolution. Based on the sequence of complementarity-determining region 3 (CDR3), the recombinant sdAb can be produced for in vitro and in vivo utilities. This method provides a general means to identify the functional measure of sdAb and its complementary epitopes and its potential applications to investigate cellular processes., (© 2021 Yin et al.)