RESUMO
Deep-learning-based classification models are increasingly used for predicting molecular properties in drug development. However, traditional classification models using the Softmax function often give overconfident mispredictions for out-of-distribution samples, highlighting a critical lack of accurate uncertainty estimation. Such limitations can result in substantial costs and should be avoided during drug development. Inspired by advances in evidential deep learning and Posterior Network, we replaced the Softmax function with a normalizing flow to enhance the uncertainty estimation ability of the model in molecular property classification. The proposed strategy was evaluated across diverse scenarios, including simulated experiments based on a synthetic dataset, ADMET predictions, and ligand-based virtual screening. The results demonstrate that compared with the vanilla model, the proposed strategy effectively alleviates the problem of giving overconfident but incorrect predictions. Our findings support the promising application of evidential deep learning in drug development and offer a valuable framework for further research.
RESUMO
Extracting knowledge from complex and diverse chemical texts is a pivotal task for both experimental and computational chemists. The task is still considered to be extremely challenging due to the complexity of the chemical language and scientific literature. This study explored the power of fine-tuned large language models (LLMs) on five intricate chemical text mining tasks: compound entity recognition, reaction role labelling, metal-organic framework (MOF) synthesis information extraction, nuclear magnetic resonance spectroscopy (NMR) data extraction, and the conversion of reaction paragraphs to action sequences. The fine-tuned LLMs demonstrated impressive performance, significantly reducing the need for repetitive and extensive prompt engineering experiments. For comparison, we guided ChatGPT (GPT-3.5-turbo) and GPT-4 with prompt engineering and fine-tuned GPT-3.5-turbo as well as other open-source LLMs such as Mistral, Llama3, Llama2, T5, and BART. The results showed that the fine-tuned ChatGPT models excelled in all tasks. They achieved exact accuracy levels ranging from 69% to 95% on these tasks with minimal annotated data. They even outperformed those task-adaptive pre-training and fine-tuning models that were based on a significantly larger amount of in-domain data. Notably, fine-tuned Mistral and Llama3 show competitive abilities. Given their versatility, robustness, and low-code capability, leveraging fine-tuned LLMs as flexible and effective toolkits for automated data acquisition could revolutionize chemical knowledge extraction.
RESUMO
INTRODUCTION: In recent years, the influence of artificial intelligence technology on clinical trials has been steadily increasing. It has brought about significant improvements in the efficiency and cost reduction of clinical trials. The objective of this scoping review is to systematically map, describe and summarise the current utilisation of artificial intelligence in recruitment and retention process of clinical trials that has been reported in research. Additionally, the review aims to identify benefits and drawbacks, as well as barriers and facilitators associated with the application of artificial intelligence in optimising recruitment and retention in clinical trials. The findings of this review will provide insights and recommendations for future development of artificial intelligence in the context of clinical trials. METHODS AND ANALYSIS: The review of relevant literature will follow the methodological framework for scoping studies provided by the Joanna Briggs Institute. A comprehensive electronic search will be conducted using the search strategy developed by the authors. Leading medical and computer science databases such as PubMed, Embase, Scopus, IEEE Xplore and Web of Science Core Collection will be searched. The search will encompass analytical observational studies, descriptive observational studies, experimental and quasi-experimental studies published in all languages, without any time limitations, which use artificial intelligence tools in the recruitment and retention process of clinical trials. The review team will screen the identified studies and import them into a dedicated electronic library specifically created for this review. Data extraction will be performed using a data charting table. ETHICS AND DISSEMINATION: Secondary data will be attained in this scoping review; therefore, no ethical approval is required. The results of the final review will be published in a peer-reviewed journal. It is expected that results will inform future artificial intelligence and clinical trials research.
Assuntos
Inteligência Artificial , Ensaios Clínicos como Assunto , Seleção de Pacientes , Humanos , Projetos de PesquisaRESUMO
BACKGROUND: Aboriginal older adults residing in remote areas have poor oral function due to inadequate access to healthcare services. Lay health advisor (LHA) strategies can fill capacity shortages of healthcare professionals in rural communities and increase population acceptance of health care or healthy behaviours. OBJECTIVE(S): To evaluate the effectiveness the LHA program on oral function and oral health-related quality of life (OHRQoL) among older adults in aboriginal community. METHODS: Participants were randomly assigned to an experimental group (EG; n = 122) and a control group (CG; n = 118). All participants performed oral exercises, and the EG received additional one-on-one 30-min lessons by an LHA over 4 weeks. Data were collected through face-to-face interview and oral examination. The generalized estimating equation model was used to analyse changes in outcomes over time. RESULTS: The EG exhibited significantly greater improvement in swallowing (ß = .63) at the 6-month follow-up and in masticatory performance (ß = .52) and pronunciation of the syllable/pa/ (ß = 2.65) at the 2-week follow-up than the CG did. The EG had a significantly lower plaque control record (ß = -.14) and plaque index (ß = -.30) at the 3-month follow-up than the CG did. Moreover, the OHRQoL was significantly increased at 6-months follow-up in the EG (p = .010). CONCLUSION: The LHA program had positive effects on chewing, swallowing and plaque control in aboriginal older adults. LHA group also experienced positive long-term effect on OHRQoL after intervention.
Assuntos
Promoção da Saúde , Qualidade de Vida , Humanos , Idoso , Assistência Odontológica , Índice de Placa Dentária , Saúde BucalRESUMO
In the field of molecular simulation for drug design, traditional molecular mechanic force fields and quantum chemical theories have been instrumental but limited in terms of scalability and computational efficiency. To overcome these limitations, machine learning force fields (MLFFs) have emerged as a powerful tool capable of balancing accuracy with efficiency. MLFFs rely on the relationship between molecular structures and potential energy, bypassing the need for a preconceived notion of interaction representations. Their accuracy depends on the machine learning models used, and the quality and volume of training data sets. With recent advances in equivariant neural networks and high-quality datasets, MLFFs have significantly improved their performance. This review explores MLFFs, emphasizing their potential in drug design. It elucidates MLFF principles, provides development and validation guidelines, and highlights successful MLFF implementations. It also addresses potential challenges in developing and applying MLFFs. The review concludes by illuminating the path ahead for MLFFs, outlining the challenges to be overcome and the opportunities to be harnessed. This inspires researchers to embrace MLFFs in their investigations as a new tool to perform molecular simulations in drug design.
Assuntos
Desenho de Fármacos , Aprendizado de Máquina , Humanos , Simulação por Computador , Estrutura MolecularRESUMO
OBJECTIVE: To evaluate the effectiveness of a clinical-based oral function intervention on oral function and care behaviours in older patients with mild dementia. METHOD: Participants were randomly assigned to the experimental group (EG) and control group (CG). Both groups received a leaflet on oral health-related knowledge, and the EG also received an oral function intervention, which was a brief one-on-one lesson concerning oral exercise and preventive oral care. Oral exercise included turning the head, pouting lips, bulging cheeks, stretching tongue, articulation exercise and salivary gland massages. A reminder phone call was made every 2 weeks. Perceived xerostomia and dysphagia, plaque index (PI), Winkel tongue-coating index (WTCI), repetitive saliva-swallowing test (RSST), oral diadochokinesis (DDK) and oral care behaviours were recorded at baseline and at 3-month follow-up. Generalised Estimating Equations (GEE) were used to analyse the indicated effects. RESULTS: The EG (n = 59) exhibited greater improvement to the CG (n = 55) in RSST [ß = 0.7; effect size (ES) = 0.45], the syllables /pa/ (ß = 3.1; ES = 0.37) and /ka/ (ß = 2.7; ES = 0.40) in oral DDK, PI (ß = -0.2; ES = 0.52) and WTCI (ß = -0.8; ES = 0.38). Moreover, the EG exhibited better preventive behaviours in regular dental visits [adjusted odds ratio (aOR) = 2.2], daily mouth cleaning frequency (aOR = 1.6) and mouth cleaning before sleep (aOR = 1.3). CONCLUSION: The brief clinical-based intervention was effective in improving the swallowing function, oral DDK and plaque control of older patients with mild dementia at 3-month follow-up.
Assuntos
Transtornos de Deglutição , Demência , Xerostomia , Idoso , Humanos , Deglutição , Saúde Bucal , Xerostomia/reabilitação , Transtornos de Deglutição/reabilitaçãoRESUMO
The commonly accepted mechanism of CO2 fixation of epoxides to cyclic carbonates catalyzed by multifunctional non-halide organocatalysts is challenged by our computational DFT-D3 study, which revealed a new polymerization-like mechanism comprising alternate epoxide and CO2 activation steps and a nested CO2 activation pathway. We investigated a recently reported CO2 coupling with epoxide reaction catalyzed by a bis-phenolic multifunctional catalyst. The predicted cis/trans product ratio is in excellent agreement with experimental results. The general applicability of the new mechanism is supported by another diamine-diacid catalyzed CO2 fixation reaction.
RESUMO
BACKGROUND: The aim of this study was to observe changes in the laryngeal vestibule under head extension or in neutral position with or without the chin lifted after anesthesia induction. METHODS: After anesthesia induction, a flexible bronchoscope was used to observe the status of the laryngeal vestibule under four positions. Then, the degree of laryngeal vestibule closure was graded. The grading scores and the incidence of severe closure under these four positions were compared. The factors that may be related to the severe closure and potential contributions of head extension and chin lifting were also analyzed. When laryngeal mask was applied, the change in the laryngeal vestibule state was observed through the laryngeal mask after placement. RESULTS: The closure scores of the laryngeal vestibule were significantly different among these four positions (P<0.001). The incidence of severe closure was highest in the neutral head position without the chin lifted. Regression analysis showed that age and snoring were associated with severe closure of the laryngeal vestibule. Both head extension and chin lifting could decrease the risk of severe closure of the laryngeal vestibule, and head extension was more important than chin lift. After laryngeal mask placement, the severely closed laryngeal vestibule can be further opened up to grades 1-2. CONCLUSIONS: The neutral position of the head after anesthesia induction may cause complete closure of the laryngeal vestibule. The best way to keep the laryngeal vestibule open is head extension with the chin lifted, while most the laryngeal vestibules will be severely closed in the neutral position without the chin lifted.
Assuntos
Máscaras Laríngeas , Laringe , Anestesia Geral , Humanos , Máscaras Laríngeas/efeitos adversos , PosturaRESUMO
The effects of pre-stretching on creep behavior, mechanical properties and microstructure during the creep aging process of Al-Cu-Li alloy were investigated. AA2195 was taken as the representative of Al-Cu-Li alloys. It is found that the total creep strain and strength property of creep aged AA2195 specimens can be improved through effective pre-stretching. Unlike with artificial aging, yield strength increased increasing by 47%. The TEM images show that the constitution of aging precipitates in the creep-aged specimens are obviously changed by pre-stretching. Precipitates in the 2% pre-stretched specimen are mainly composed of T1 phase, while a great amount of θ' phase accompanied with a few T1 phase were found in the non-pre-stretched specimen. Moreover, pre-stretching introduces many dislocations which benefit the creep deformation, but the increasing dislocation density also accelerates the nucleation and growth of the precipitates as well. The premature T1 phase has a great blocking effect to the dislocation motion, creating a lower decrease rate but a longer duration in the early creep stage. Except for the initial dislocations, the dislocation motion in the creep aging process is also a favorable factor to precipitate the T1 phase.