Gut microbiome-based disease diagnosis holds significant promise but remains challenging due to the high data dimensionality, typically small sample sizes, and the necessity of incorporating biological knowledge. Due to these challenges, traditional machine learning approaches often tend to overfit the data and fail to capture true biological relationships, resulting in inaccurate diagnoses. To fill in the gap, we propose a two-phase, knowledge-guided large language model (LLM) framework for disease diagnosis that integrates biomedical expertise with in-context learning. In Phase 1, an LLM is employed to identify disease-associated taxa from hundreds of microbial families and to infer their biological relationships with the disease outcome. This process reduces the feature space dimensionality through biologically-informed feature selection and acquires essential domain knowledge. In Phase 2, we employ few-shot prompting to guide the LLM in disease outcome classification based on the domain knowledge acquired in Phase 1. Thanks to the universal applicability of LLM and our two-phase approach, this is a generic framework that can be applied to a wide range of microbiome-based disease diagnostic tasks. We demonstrate the superiority of our framework using inflammatory bowel disease (IBD) as a representative case study, where our approach achieves an accuracy of 73.91%, significantly outperforming an optimized XGBoost classifier. Overall, our knowledge-guided framework provides a powerful and generalizable strategy for leveraging LLMs in microbiome-based disease diagnosis, and opens a new avenue for disease diagnosis in the era of LLM.

SUMMARY: Identification of balances of bacterial taxa in relation to continuous and dichotomous outcomes is an increasingly frequent analytic objective in microbiome profiling experiments. SurvBal enables the selection of balances in relation to censored survival or time-to-event outcomes which are of considerable interest in many biomedical studies. The most commonly used survival models-the Cox proportional hazards and parametric survival models are included in the package, which are used in combination with step-wise selection procedures to identify the optimal associated balance of microbiome, i.e. the ratio of the geometric means of two groups of taxa's relative abundances. AVAILABILITY AND IMPLEMENTATION: The SurvBal R package and Shiny app can be accessed at https://github.com/yinglia/SurvBal and https://yinglistats.shinyapps.io/shinyapp-survbal/.

BACKGROUND: Microbial communities are known to be closely related to many diseases, such as obesity and HIV, and it is of interest to identify differentially abundant microbial species between two or more environments. Since the abundances or counts of microbial species usually have different scales and suffer from zero-inflation or over-dispersion, normalization is a critical step before conducting differential abundance analysis. Several normalization approaches have been proposed, but it is difficult to optimize the characterization of the true relationship between taxa and interesting outcomes. RESULTS: To avoid the challenge of picking an optimal normalization and accommodate the advantages of several normalization strategies, we propose an omnibus approach. Our approach is based on a Cauchy combination test, which is flexible and powerful by aggregating individual p values. We also consider a truncated test statistic to prevent substantial power loss. We experiment with a basic linear regression model as well as recently proposed powerful association tests for microbiome data and compare the performance of the omnibus approach with individual normalization approaches. Experimental results show that, regardless of simulation settings, the new approach exhibits power that is close to the best normalization strategy, while controling the type I error well. CONCLUSIONS: The proposed omnibus test releases researchers from choosing among various normalization methods and it is an aggregated method that provides the powerful result to the underlying optimal normalization, which requires tedious trial and error. While the power may not exceed the best normalization, it is always much better than using a poor choice of normalization.

MOTIVATION: Most existing microbiome association analyses focus on the association between microbiome and conditional mean of health or disease-related outcomes, and within this vein, vast computational tools and methods have been devised for standard binary or continuous outcomes. However, these methods tend to be limited either when the underlying microbiome-outcome association occurs somewhere other than the mean level, or when distribution of the outcome variable is irregular (e.g. zero-inflated or mixtures) such that conditional outcome mean is less meaningful. We address this gap by investigating association analysis between microbiome compositions and conditional outcome quantiles. RESULTS: We introduce a new association analysis tool named MiRKAT-IQ within the Microbiome Regression-based Kernel Association Test framework using Integrated Quantile regression models to examine the association between microbiome and the distribution of outcome. For an individual quantile, we utilize the existing kernel machine regression framework to examine the association between that conditional outcome quantile and a group of microbial features (e.g. microbiome community compositions). Then, the goal of examining microbiome association with the whole outcome distribution is achieved by integrating all outcome conditional quantiles over a process, and thus our new MiRKAT-IQ test is robust to both the location of association signals (e.g. mean, variance, median) and the heterogeneous distribution of the outcome. Extensive numerical simulation studies have been conducted to show the validity of the new MiRKAT-IQ test. We demonstrate the potential usefulness of MiRKAT-IQ with applications to actual biological data collected from a previous microbiome study. AVAILABILITY AND IMPLEMENTATION: R codes to implement the proposed methodology is provided in the MiRKAT package, which is available on CRAN. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Successful prediction of clinical outcomes facilitates tailored diagnosis and treatment. The microbiome has been shown to be an important biomarker to predict host clinical outcomes. Further, the incorporation of microbial phylogeny, the evolutionary relationship among microbes, has been demonstrated to improve prediction accuracy. We propose a phylogeny-driven deep neural network (PhyNN) and develop an ensemble method, DeepEn-Phy, for host clinical outcome prediction. The method is designed to optimally extract features from phylogeny, thereby take full advantage of the information in phylogeny while harnessing the core principles of phylogeny (in contrast to taxonomy). We apply DeepEn-Phy to a real large microbiome data set to predict both categorical and continuous clinical outcomes. DeepEn-Phy demonstrates superior prediction performance to existing machine learning and deep learning approaches. Overall, DeepEn-Phy provides a new strategy for designing deep neural network architectures within the context of phylogeny-constrained microbiome data.

Objective: While depression and anxiety are common mental health issues, only a small segment of the population has access to standard one-on-one treatment. The use of smartphone apps can fill this gap. An app recommender system may help improve user engagement of these apps and eventually symptoms. Methods: IntelliCare was a suite of apps for depression and anxiety, with a Hub app that provided app recommendations aiming to increase user engagement. This study captured the records of 8057 users of 12 apps. We measured overall engagement and app-specific usage longitudinally by the number of weekly app sessions ("loyalty") and the number of days with app usage ("regularity") over 16 weeks. Hub and non-Hub users were compared using zero-inflated Poisson regression for loyalty, linear regression for regularity, and Cox regression for engagement duration. Adjusted analyses were performed in 4561 users for whom we had baseline characteristics. Impact of Hub recommendations was assessed using the same approach. Results: When compared to non-Hub users in adjusted analyses, Hub users had a lower risk of discontinuing IntelliCare (hazard ratio = 0.67, 95% CI, 0.62-0.71), higher loyalty (2- to 5-fold), and higher regularity (0.1-0.4 day/week greater). Among Hub users, Hub recommendations increased app-specific loyalty and regularity in all 12 apps. Discussion/Conclusion: Centralized app recommendations increase overall user engagement of the apps, as well as app-specific usage. Further studies relating app usage to symptoms can validate that such a recommender improves clinical benefits and does so at scale.