Predictive Power Research Articles

Multidimensional house price prediction with SOTA RNNs

This paper introduces insights into the Turkish real estate market, which can be generalized globally. It primarily aims to find the best forecasting algorithms for the housing price index and compare their prediction performance over three, six, nine, and twelve months ahead by using recurrent neural networks (RNN) with a comparison of out-of-sample predicting power of econometrical models. For these purposes, we employ three RNN architectures in twenty-four settings, revealing that certain RNN architectures are the best predictors in forecasting the Turkish real housing price index. The RNN architectures outperform traditional econometric models; however, the more months forecasted, the lower the prediction power. The lagged values of the price-to-rent ratio, real rents, and the lagged USDTRY values contribute more than the other predictors in forecasting the real housing price index. The outcomes suggest that stocks, real estate investment trusts, and gold are neither complementary nor competing financial instruments since housing is an illiquid asset.

Frailty and Overall Survival of Older Patients Undergoing Radiotherapy for Head and Neck Cancer: A Prospective Analysis

Background/Objectives: Over 60% of diagnoses of head and neck squamous cell carcinoma (HNSCC) occur in patients aged over 65. The benefits of radiotherapy (RT) combined with chemotherapy remain controversial in geriatric patients. Frailty stratification is underutilized in clinical practice despite evidence suggesting its predictive power. We aimed to assess the impact of pre-treatment frailty stratification on overall survival (OS) and acute radiation-related toxicity in a cohort of older patients with HNSCC. Methods: A prospective cohort of 117 patients aged ≥65 years with newly diagnosed HNSCC was enrolled between 2017 and 2022. Patients received RT with or without chemotherapy. Frailty was assessed using the 40-item Frailty Index (FI) as part of a comprehensive geriatric assessment. The primary outcomes were OS and acute RT toxicity, defined as a grade ≥ 3 based on CTCAE 5.0. Cox proportional hazard models and logistic regression models, adjusted for age, tumor stage, and chemotherapy, were used. Results: After a median follow-up of 819 days, FI (HR 1.478, 95% CI 1.182–1.848, p < 0.001) was an independent predictor of OS. Grade ≥ 3 toxicities were observed in 38% of patients, but frailty was not significantly associated with toxicity. A sub-analysis of oropharyngeal carcinoma patients demonstrated improved OS in HPV-positive patients (HR 0.231, 95% CI 0.078–0.686, p = 0.008). Conclusions: the standardized introduction of frailty assessment in the evaluation of older HNSCC patients holds great promise for predicting long-term survival. This approach could be crucial in tailoring treatment protocols that account for the biological heterogeneity of older adults.

Assessing the Effect of Rewards on Academic Staff Performance in Uganda's Public Universities

The persistent decline in academic staff performance at Uganda's public universities prompted this study to investigate the effect of rewards on their performance. Using an exploratory sequential research design, data were collected from 350 academic staff through self-administered questionnaires and analyzed using descriptive statistics and linear regression, alongside interviews with 13 academic staff, which were analyzed using content analysis. The findings reveal that while rewards are a statistically significant predictor of performance, their influence is modest, accounting for 7.6% of the variation in teaching performance, 12.5% in research performance, and 9.6% in community service performance. These results suggest that while rewards positively impact academic performance, their limited predictive power indicates that other factors play a more significant role. The qualitative findings supported the quantitative data, highlighting the importance of additional influences such as professional development, workplace environment, and institutional support. As a result, it is recommended that public universities not only strengthen reward systems but also adopt a comprehensive approach that addresses these other factors. This holistic strategy will help create a more conducive academic environment and ultimately enhance the overall performance of academic staff

N-glycan as new potential biomarker for predicting treatment response in patients with type 2 diabetes mellitus.

To investigate the N-glycans related to the metformin efficacy in patients with type 2 diabetes mellitus (T2DM). We enrolled 141healthy controls and 195 newly diagnosed T2DM patients treated with metformin for 3 months. Serum N-glycan profile was determined by DNA sequencer - assisted fluorophore-assisted carbohydrate electrophoresis (DSA-FACE). The N-glycan model was established by logistic regression analysis. Receiver operating characteristic curve (ROC) analysis was used to analyze the predictive power of the N-glycan model for metformin efficacy. The abundances of several N-glycans in serum of T2DM patients at baseline were significantly different from those of healthy controls and tended to recover the N-glycan of controls after 3 months treatment. Serum N-glycans changes were more significant in the good response group (FPG <7 mmol/L) after metformin treatment. In addition, the abundance of peak9 at baseline had an opposite tendency between HbA1c increased and decreased groups post-treatment, which could be a biomarker for predicting metformin efficacy. Peak9 combined with other 11 N-glycans at baseline was used to establish the predictive model to distinguish non-response from response patients (AUROC = 0.780, sensitivity = 70.6% and specificity = 77.5%). Serum N-glycans may have potential value as biomarkers for indicating the efficacy of metformin.

Uniform in Number of Neighbor Consistency and Weak Convergence of k-Nearest Neighbor Single Index Conditional Processes and k-Nearest Neighbor Single Index Conditional U-Processes Involving Functional Mixing Data

U-statistics are fundamental in modeling statistical measures that involve responses from multiple subjects. They generalize the concept of the empirical mean of a random variable X to include summations over each m-tuple of distinct observations of X. W. Stute introduced conditional U-statistics, extending the Nadaraya–Watson estimates for regression functions. Stute demonstrated their strong pointwise consistency with the conditional expectation r(m)(φ,t), defined as E[φ(Y1,…,Ym)|(X1,…,Xm)=t] for t∈Xm. This paper focuses on estimating functional single index (FSI) conditional U-processes for regular time series data. We propose a novel, automatic, and location-adaptive procedure for estimating these processes based on k-Nearest Neighbor (kNN) principles. Our asymptotic analysis includes data-driven neighbor selection, making the method highly practical. The local nature of the kNN approach improves predictive power compared to traditional kernel estimates. Additionally, we establish new uniform results in bandwidth selection for kernel estimates in FSI conditional U-processes, including almost complete convergence rates and weak convergence under general conditions. These results apply to both bounded and unbounded function classes, satisfying certain moment conditions, and are proven under standard Vapnik–Chervonenkis structural conditions and mild model assumptions. Furthermore, we demonstrate uniform consistency for the nonparametric inverse probability of censoring weighted (I.P.C.W.) estimators of the regression function under random censorship. This result is independently valuable and has potential applications in areas such as set-indexed conditional U-statistics, the Kendall rank correlation coefficient, and discrimination problems.

Uncertainty, Skewness, and the Business Cycle Through the MIDAS Lens

ABSTRACTWe employ a mixed‐frequency quantile regression approach to model the time‐varying conditional distribution of the US real GDP growth rate. We show that monthly information on financial conditions improves the predictive power of an otherwise quarterly‐only model. We combine selected quantiles of the estimated conditional distribution to produce novel measures of uncertainty and skewness. Embedding these measures in a VAR framework, we show that unexpected changes in uncertainty are associated with an increase in (left) skewness and a downturn in real activity. Business cycle effects are significantly downplayed if we consider a quarterly‐only quantile regression model. We find the endogenous response of skewness to substantially amplify the recessionary effects of uncertainty shocks. Finally, we construct a monthly frequency version of our uncertainty measure and document the robustness of our findings.

Predicting ICU Admission for Hospitalized COVID-19 Patients by Artificial Neural Network Combined with Elastic Net

Background: Machine learning models could assist physicians in identifying high-risk COVID-19 patients. This study aimed to predict the Intensive Care Unit (ICU) admission in COVID-19 hospitalized patients by the Artificial Neural Network (ANN) model combined with Elastic Net algorithm. Methods: In this prospective study, the data of 139 COVID-19 patients admitted to Imam Reza Hospital in Tabriz between 20 March and 5 April 2020, were analyzed. The Elastic Net method was used to choose features with high importance. The chosen variables were standardized and ANN was fitted to the data with one hidden layer based on the descending gradient algorithm. To validate the model, the training and test group method with a ratio of 70 to 30 was used. The model’s predictive power was reported by calculating the overall accuracy, sensitivity, specificity, and Area Under the ROC curve (AUC). Results: According to the results of the Elastic Net, the ANN model was constructed based on age, sex, body mass index, diabetes mellitus status, history of heart disease, and vital signs including systolic and diastolic blood pressure, saturated oxygen level, pulse rate, respiration rate, and body temperature. The overall accuracy of this model was 93.15%, sensitivity 80%, specificity 95.8%, and AUC 0.90. Saturated oxygen level, pulse rate, and age were the most important and predictive variables. Conclusion: In the investigated sample of patients admitted with COVID-19, the fitted ANN model had acceptable performance to predict the ICU admission. This finding could be useful for physicians and policy makers.

The systemic immune-inflammation index and systemic inflammation response index are useful for predicting mortality in patients with diabetic nephropathy

BackgroundThis study investigated the correlation between the systemic immune-inflammation index (SII) and the systemic inflammation response index (SIRI) and all-cause, cardiovascular, and kidney disease mortality in patients with diabetic nephropathy (DN). It aimed to provide a new predictive assessment tool for the clinic and a scientific basis for managing inflammation in DN.MethodsThe data utilized in this study were obtained from the National Health and Nutrition Examination Survey (NHANES) database, spanning 1999 to 2018. A total of 2641 patients diagnosed with DN were included in the analysis. The association between SII and SIRI levels and mortality in patients with DN was investigated using multivariate Cox proportional risk regression models. These relationships were further validated by Kaplan-Meier survival curves and restricted cubic spline (RCS) modeling, and subgroup analyses were performed to explore the heterogeneity among different characteristic subgroups.ResultsThe multivariate Cox regression analysis indicated that SII and SIRI levels were independently associated with all-cause mortality and cardiovascular mortality in patients with DN. SIRI levels were found to be an independently associated factor with kidney disease mortality in patients with DN. Patients in the highest quartile of SII and SIRI exhibited a 1.49-fold and 1.62-fold increased risk of all-cause mortality, respectively, compared to patients in the lowest quartile. The risk of cardiovascular mortality was 1.31 and 1.73 times higher than that in patients in the lowest quartile, respectively. The risk of kidney disease mortality in patients in the highest quartile of SIRI was 2.74 times higher than that in patients in the lowest quartile. Kaplan-Meier survival curve and RCS analyses further confirmed the positive association between SII and SIRI and mortality and a significant nonlinear relationship between SII and all-cause mortality. The SII and SIRI indices offer incremental value in model predictive power for mortality in patients with DN. Subgroup analyses demonstrated that the correlation between SII and SIRI and mortality risk was stable but heterogeneous across different subgroups.ConclusionSII and SIRI can be utilized as biomarkers for forecasting the likelihood of all-cause and cardiovascular mortality in patients with DN.

Factors affecting quality of life among older adults with hypertension in Wenzhou, China: A cross-sectional study.

In China, the incidence of hypertension rises significantly with age, resulting in a markedly reduced quality of life (QoL) among older patients compared to the general population. Therefore, it is essential for healthcare providers, particularly nurses, to identify the predictive factors that influence QoL in this demographic. This study aimed to describe the QoL levels and investigate the predictive power of perceived health status, self-care behavior, and social support on QoL among older patients with hypertension in Wenzhou, China. A cross-sectional study was conducted with 131 patients with hypertension aged 60 and above, visiting the Cardiovascular Outpatient Department of The Second Affiliated Hospital of Wenzhou Medical University. Data were collected using validated instruments between November and December 2022 and analyzed using descriptive statistics and stepwise multiple regression. The overall QoL was moderate (Mean = 75.52, SD = 5.86). Self-care behavior (β = 0.421, p <0.001), social support (β = 0.416, p <0.001), and perceived health status (β = -0.170, p <0.001) were significant predictors of QoL, collectively explaining 82.7% of the variance. The findings highlight the importance of self-care behavior, social support, and perceived health status as critical factors influencing QoL among older patients with hypertension. Nurses and other healthcare providers should focus on enhancing these areas through targeted education and support initiatives to improve the overall well-being of this vulnerable population.

Dissecting the Tumor Microenvironment (TME) to Decipher New Immunotherapy Targets by Using Artificial Intelligence

Cancer remains one of the leading causes of death worldwide, second only to cardiovascular diseases. Standard cancer treatments, such as chemotherapy, radiation, and surgery, are effective for primary tumors but often fail to eliminate disseminated tumor cells responsible for metastasis. This limitation underscores the necessity for advanced therapeutic strategies, leading to the rise of immunotherapy, which leverages the immune system to combat cancer. Despite its promise, immunotherapy faces challenges, including variable patient responses and immune-related toxicities, complicating the prediction of treatment efficacy. Here, artificial intelligence (AI) emerges as a vital tool that can enhance the precision and effectiveness of immunotherapy by analyzing the intricate tumor microenvironment (TME). This paper explores the limitations of current immunotherapies and examines how AI can address these challenges. It discusses the TME’s role in shaping immune responses, highlighting how understanding its complexities can improve predictive power and treatment outcomes. Furthermore, we address the limitations of AI in cancer research and propose future directions for its integration into clinical practice, with the potential to revolutionize personalized cancer therapy and improve overall patient care.

Exposing factors influencing Korean leisure life satisfaction through machine learning techniques

This study examines factors influencing leisure life satisfaction (LLS) through machine learning techniques based on the data from the 2019 National Leisure Activity Survey in Korea. The results show that using machine learning techniques in identifying LLS influencing factors improves predictive power and helps detect effective leisure interventions. We also provide insights into the factors influencing LLS by standardizing activity measures based on leisure type and examining differences in resource accessibility and experiences across groups. The findings suggest that a diverse and balanced leisure repertoire is associated with greater levels of LLS, particularly in active leisure and social activities. However, the relationship between the repertoire of passive leisure activities and LLS is negative, suggesting that the optimal point for leisure activities lies found between various leisure experiences and limited resources. Leisure resource availability, such as expenditure, time, facilities, and interpersonal factors, may affect LLS, but varies by age. The results provide new insights and more accurate models of the factors influencing LLS and their complex relationships.

Research on traffic congestion prediction based on analyzable machine learning

As motor vehicles proliferate around the world, traffic congestion increases, hindering travel, polluting the environment and hampering economic growth. Accurate congestion prediction is the key to optimizing traffic flow. In this study, the interaction between different traffic modes and congestion degree is deeply investigated, and a high-fidelity prediction model is established. The study analyses monthly, weekly, daily and hourly multi-modal traffic and congestion data and identifies congestion patterns. In this paper, Pearson's coefficient is applied to identify the key influencing factors, and congestion is predicted by KNN, logistic regression, decision tree and random forest models. The random forest model reaches 99.88% accuracy after 5-fold cross-validation and grid search optimisation. Further analysis using the Random Forest's feature importance scores revealed that traffic volume was the most important predictor of congestion, while time factors such as time and date played a smaller role. This study not only establishes a robust congestion prediction model, but also emphasizes the importance of traffic volume in congestion prediction. The findings provide valuable insights for traffic management agencies to take proactive measures to alleviate traffic congestion. By harnessing the predictive power of random forest models, these agencies can increase the efficiency of traffic management, improve the overall travel experience, and contribute to sustainable urban development. This study also lays a solid foundation for future research in the field of traffic congestion prediction and management.

The Predictive Value of Fibrinogen-to-Albumin Ratio for Predicting Intravenous Immunoglobulin Resistance in Kawasaki Disease: A Prospective Cohort Study.

Predicting resistance to intravenous immunoglobulin (IVIG) in the treatment of Kawasaki disease (KD) remains a focus of research. Fibrinogen and albumin in systemic inflammation play an important role. This study aims to investigate the predictive value of fibrinogen to albumin ratio (FAR) for initial IVIG resistance in patients with KD. The study prospectively recruited 962 patients with KD between July 2015 and June 2022. The serum characteristics of the two groups were compared by comparing fibrinogen and albumin, as well as other laboratory and clinical data between the IVIG-responsive and IVIG-resistant groups. Multivariate logistic regression was used to explore the relationship between FAR and IVIG resistance. Receiver operating characteristic (ROC) curves were used to determine the effectiveness of FAR in predicting initial IVIG resistance. Our results demonstrated that IVIG-resistant patients had significantly higher fibrinogen levels (603.35 ± 99.00 mg/L), FAR (17.30 ± 3.31), and lower albumin (35.47 ± 5.24 g/L) compared to IVIG-responsive patients (fibrinogen 572.35 ± 145.75 mg/L; FAR 15.08 ± 4.32; albumin 38.52 ± 4.55 g/L). 15.20 was the best cut-off value of FAR for predicting initial IVIG resistance. The sensitivity was 72.5%, the specificity was 51.3%, the positive predictive value was 91.8%, and the negative predictive value was 20.0%. Multivariate logistic regression analysis, found that FAR was an independent predictor of initial IVIG resistance in KD children. The FAR was an independent risk factor for initial IVIG resistance, its predictive power for initial IVIG resistance exceeded that of albumin and fibrinogen alone. FAR may not be suitable as a single marker but might serve as a complementary laboratory marker to accurately predict initial IVIG resistance in KD.

Effect of Water Networks On Ligand Binding: Computational Predictions vs Experiments.

Rational drug design focuses on the explanation and prediction of complex formation between therapeutic targets and small-molecule ligands. As a third and often overlooked interacting partner, water molecules play a critical role in the thermodynamics of protein-ligand binding, impacting both the entropy and enthalpy components of the binding free energy and by extension, on-target affinity and bioactivity. The community has realized the importance of binding site waters, as evidenced by the number of computational tools to predict the structure and thermodynamics of their networks. However, quantitative experimental characterization of relevant protein-ligand-water systems, and consequently the validation of these modeling methods, remains challenging. Here, we investigated the impact of solvent exchange from light (H2O) to heavy water (D2O) to provide complete thermodynamic profiling of these ternary systems. Utilizing the solvent isotope effects, we gain a deeper understanding of the energetic contributions of various components. Specifically, we conducted isothermal titration calorimetry experiments on trypsin with a series of p-substituted benzamidines, as well as carbonic anhydrase II (CAII) with a series of aromatic sulfonamides. Significant differences in binding enthalpies found between light vs heavy water indicate a substantial role of the binding site water network in protein-ligand binding. Next, we challenged two conceptually distinct modeling methods, the grid-based WaterFLAP and the molecular dynamics-based MobyWat, by predicting and scoring relevant water networks. The predicted water positions accurately reproduce those in available high-resolution X-ray and neutron diffraction structures of the relevant protein-ligand complexes. Estimated energetic contributions of the identified water networks were corroborated by the experimental thermodynamics data. Besides providing a direct validation for the predictive power of these methods, our findings confirmed the importance of considering binding site water networks in computational ligand design.

Identification of inflammation related gene signatures for bladder cancer prognosis prediction

Early diagnosis and treatment of bladder cancer are crucial, and since inflammation plays a role in all stages of bladder cancer, this study aims to develop a model based on inflammation-related genes to accurately predict patient prognosis. The data were initially processed through differential analysis and prognostic correlation analysis, then a Least absolute shrinkage and selection operator (LASSO) regression model was constructed by M-cohort and a nomogram was designed to increase the model readability. The T-cohort was used for internal validation, with the GSE32894 and Imvigor210 cohorts used as external data to verify the model’s accuracy. The model’s predictive ability was verified for the prognosis of patients of different ages, gender, tumor stage, and tumour grade. The GSE3167, GSE13507 and GeneExpression Profiling Interactive Analysis (GEPIA) datasets and Human Protein Atlas (HPA) database were used to verify the expression of the inflammation-related genes, which were confirmed by real-time Polymerase Chain Reaction (PCR). A comprehensive analysis of the model’s inflammation-related genes, Gene Set Enrichment Analysis (GSEA), Gene Set Variation Analysis (GSVA) enrichment analysis, and immune-related analysis were also performed. Both internal and external data validations confirmed that the developed model can accurately predict the prognosis across different patient populations. Hierarchical validation results demonstrated that the model’s predictive power is reliable for various patient stratifications. The expression of inflammation-related genes was consistent across The Cancer Genome Atlas (TCGA) database, GSE3167 dataset, GSE13507 dataset, Gene Expression Profiling Interactive Analysis (GEPIA) database, and the Human Protein Atlas (HPA) database, and was further validated by real-time Polymerase Chain Reaction (PCR). Pathway enrichment analysis indicated that patients in the high-risk (H-risk) group exhibited a variety of tumors. Meanwhile, patients in the low-risk (L-risk) group may be candidates for immunotherapy, whereas those in the high-risk group are more likely to benefit from chemotherapy. The model of inflammation-related genes can accurately predict bladder cancer patient prognosis, with MEST, FASN, KRT6B, and RGS2 anticipated to become new prognostic bladder cancer markers.

Deep neural networks for endemic measles dynamics: Comparative analysis and integration with mechanistic models.

Measles is an important infectious disease system both for its burden on public health and as an opportunity for studying nonlinear spatio-temporal disease dynamics. Traditional mechanistic models often struggle to fully capture the complex nonlinear spatio-temporal dynamics inherent in measles outbreaks. In this paper, we first develop a high-dimensional feed-forward neural network model with spatial features (SFNN) to forecast endemic measles outbreaks and systematically compare its predictive power with that of a classical mechanistic model (TSIR). We illustrate the utility of our model using England and Wales measles data from 1944-1965. These data present multiple modeling challenges due to the interplay between metapopulations, seasonal trends, and nonlinear dynamics related to demographic changes. Our results show that while the TSIR model yields similarly performant short-term (1 to 2 biweeks ahead) forecasts for highly populous cities, our neural network model (SFNN) consistently achieves lower root mean squared error (RMSE) across other forecasting windows. Furthermore, we show that our spatial-feature neural network model, without imposing mechanistic assumptions a priori, can uncover gravity-model-like spatial hierarchy of measles spread in which major cities play an important role in driving regional outbreaks. We then turn our attention to integrative approaches that combine mechanistic and machine learning models. Specifically, we investigate how the TSIR can be utilized to improve a state-of-the-art approach known as Physics-Informed-Neural-Networks (PINN) which explicitly combines compartmental models and neural networks. Our results show that the TSIR can facilitate the reconstruction of latent susceptible dynamics, thereby enhancing both forecasts in terms of mean absolute error (MAE) and parameter inference of measles dynamics within the PINN. In summary, our results show that appropriately designed neural network-based models can outperform traditional mechanistic models for short to long-term forecasts, while simultaneously providing mechanistic interpretability. Our work also provides valuable insights into more effectively integrating machine learning models with mechanistic models to enhance public health responses to measles and similar infectious disease systems.

Microvascular disease and severe COVID-19 outcomes in UKBiobank participants with diabetes.

Early in the COVID-19 pandemic, evidence emerged suggesting that people with diabetic retinopathy (DR) or other microvascular diseases had greater risk of severe short-term outcomes. This study evaluated longer-term outcomes, providing more generalisable evidence. We identified a cohort of UKBiobank participants with diabetes and retrieved their diagnostic codes for a variety of microvascular diseases, complications of diabetes and systemic comorbidities. We investigated relationships between diagnoses and the study outcome: admission to Critical Care or death from COVID-19, taking age, sex and diabetes duration into account. We tested relationships, adding baseline covariates and weighting diagnostic codes according to their recency prior to COVID-19 diagnosis. In univariate analyses, DR (OR: 1·519, p = 0·016) and microvascular disease (OR: 2·001, p = 0·000) were associated with greater risk of the outcome. In multivariate analyses, as expected, respiratory disease was most strongly associated with the study outcome, microvascular disease second. Adjusting analyses by number of admissions (general health proxy) and weighted diagnostic coding (comorbidity severity at COVID-19 diagnosis indicator), did not improve predictive power of the model. The presence of microvascular disease in routinely-collected healthcare data predicts risk of COVID-19 severe outcomes, independently of general health, in a cohort of people with diabetes.

Transdiagnostic factors predicting prescription opioid-use disorder severity: A 12-month prospective study in patients on long-term opioid therapy for chronic pain

BackgroundResearch has suggested that transdiagnostic factors related to reward, cognitive, and regulatory processes are involved in addictive behaviors and the experience of pain. However, studies of Prescription Opioid-Use Disorder (POUD) in a chronic pain population are scarce. This study aimed to analyze the predictive power of anticipatory pleasure experience, obsessive-compulsive behavior, cognitive control, emotion dysregulation, and sleep on POUD severity in chronic pain patients on long-term opioid therapy. MethodsA three-wave prospective study was conducted in a sample of 67 patients. Individual interviews were held to collect self-reported data on transdiagnostic factors, POUD, and sociodemographic and clinical variables. Statistical analysis included repeated measures multinomial mixed models, unadjusted and covariate-adjusted. ResultsRates of moderate-severe POUD significantly increased at 6- and 12-month follow-ups compared to initial assessment. The inability to experience anticipatory pleasure (OR [95 %CI] = 0.93 [0.89–0.98]), emotion dysregulation (OR = 1.06 [1.03–1.10]), and poorer sleep quality (OR = 1.25 [1.07–1.45]) predicted moderate-severe POUD even in the adjusted models (p< 0.05). In contrast, obsessive-compulsive behavior (OR = 1.02 [0.99–1.06]) was no longer significant after adjustment for covariates (p> 0.05). In addition, cognitive control was the only factor that predicted both mild (OR = 0.96 [0.93–0.99]) and moderate-severe (OR = 0.94 [0.90–0.98]) disorder (p< 0.05). Furthermore, when all transdiagnostic factors were included in a single model, cognitive control and anticipatory pleasure experience emerged as the strongest predictors of POUD severity, respectively (p< 0.05). ConclusionsThe identification of key transdiagnostic factors related to POUD severity allows for a more specific profiling of patients at increased risk, potentially leading to improved treatment targets for chronic pain population on opioid therapy.

Predictive Power of Polygenic Risk Scores for Intraocular Pressure or Vertical Cup-Disc Ratio

Early detection of glaucoma is essential to timely monitoring and treatment, and primary open-angle glaucoma risk can be assessed by measuring intraocular pressure (IOP) or optic nerve head vertical cup-disc ratio (VCDR). Polygenic risk scores (PRSs) could provide a link between genetic effects estimated from genome-wide association studies (GWASs) and clinical applications to provide estimates of an individual's genetic risk by combining many identified variants into a score. To construct IOP and VCDR PRSs with clinically relevant predictive power. This genetic association study evaluated the PRSs for 6959 of 51 338 individuals in the Canadian Longitudinal Study on Aging (CLSA; 2010 to 2015 with data from 11 centers in Canada) and 4960 of 5107 individuals the community-based Busselton Healthy Aging Study (BHAS; 2010 to 2015 in Busselton, Western Australia) with an artificial intelligence grading approach used to obtain precise VCDR estimates for the CLSA dataset. Data for approximately 500 000 individuals in UK Biobank from 2006 to 2010 were used to validate the power of the PRS. Data were analyzed from June to November 2023. IOP and VCDR PRSs and phenotypic variance (R2) explained by each PRS. Participants in CLSA were aged 45 to 85 years; those in BHAS, 46 to 64 years; and those in UK Biobank, 40 to 69 years. The VCDR PRS explained 22.0% (95% CI, 20.1-23.9) and 19.7% (95% CI, 16.3-23.3) of the phenotypic variance in VCDR in CLSA and BHAS, respectively, while the IOP PRS explained 12.9% (95% CI, 11.3-14.6) and 9.6% (95% CI, 8.1-11.2) of phenotypic variance in CLSA and BHAS IOP measurements. The VCDR PRS variance explained 5.2% (95% CI, 3.6-7.1), 12.1% (95% CI, 7.5-17.5), and 14.3% (95% CI, 9.3-19.9), and the IOP PRS variance explained 2.3% (95% CI, 1.5-3.3), 3.2% (95% CI, 1.3-5.8), and 7.5% (95% CI, 6.2-8.9) (P < .001) across African, East Asian, and South Asian populations, respectively. VCDR and IOP PRSs derived using a large recently published multitrait GWAS exhibited validity across independent cohorts. The findings suggest that an IOP PRS has the potential to identify individuals who may benefit from more intensive IOP-lowering treatments, which could be crucial in managing glaucoma risk more effectively. Individuals with a high VCDR PRS may be at risk of developing glaucoma even if their IOP measures fall within the normal range, suggesting that these PRSs could help in early detection and intervention, particularly among those who might otherwise be considered at low risk based on IOP alone.

Peptide-aware chemical language model successfully predicts membrane diffusion of cyclic peptides.

Language modeling applied to biological data has significantly advanced the prediction of membrane penetration for small molecule drugs and natural peptides. However, accurately predicting membrane diffusion for peptides with pharmacologically relevant modifications remains a substantial challenge. Here, we introduce PeptideCLM, a peptide-focused chemical language model capable of encoding peptides with chemical modifications, unnatural or non-canonical amino acids, and cyclizations. We assess this model by predicting membrane diffusion of cyclic peptides, demonstrating greater predictive power than existing chemical language models. Our model is versatile and can be extended beyond membrane diffusion predictions to other target values. Its advantages include the ability to model macromolecules using chemical string notation, a largely unexplored domain, and a simple, flexible architecture that allows for adaptation to any peptide or other macromolecule dataset.