Leveraging permutation testing to assess confidence in positive-unlabeled learning applied to high-dimensional biological datasets.

BACKGROUND: Compared to traditional supervised machine learning approaches employing fully labeled samples, positive-unlabeled (PU) learning techniques aim to classify "unlabeled" samples based on a smaller proportion of known positive examples. This more challenging modeling goal reflects many real-world scenarios in which negative examples are not available-posing direct challenges to defining prediction accuracy and robustness. While several studies have evaluated predictions learned from only definitive positive examples, few have investigated whether correct classification of a high proportion of known positives (KP) samples from among unlabeled samples can act as a surrogate to indicate model quality. RESULTS: In this study, we report a novel methodology combining multiple established PU learning-based strategies with permutation testing to evaluate the potential of KP samples to accurately classify unlabeled samples without using "ground truth" positive and negative labels for validation. Multivariate synthetic and real-world high-dimensional benchmark datasets were employed to demonstrate the suitability of the proposed pipeline to provide evidence of model robustness across varied underlying ground truth class label compositions among the unlabeled set and with different proportions of KP examples. Comparisons between model performance with actual and permuted labels could be used to distinguish reliable from unreliable models. CONCLUSIONS: As in fully supervised machine learning, permutation testing offers a means to set a baseline "no-information rate" benchmark in the context of semi-supervised PU learning inference tasks-providing a standard against which model performance can be compared.

Permutation-Based Inference for Function-on-Scalar Regression With an Application in PET Brain Imaging.

The density of various proteins throughout the human brain can be studied through the use of positron emission tomography (PET) imaging. We report here on data from a study of serotonin transporter (5-HTT) binding. While PET imaging data analysis is most commonly performed on data that are aggregated into several discrete a priori regions of interest, in this study, primary interest is on measures of 5-HTT binding potential that are made at many locations along a continuous anatomically defined tract, one that was chosen to follow serotonergic axons. Our goal is to characterize the binding patterns along this tract and also to determine how such patterns differ between control subjects and depressed patients. Due to the nature of our data, we utilize function-on-scalar regression modeling to make optimal use of our data. Inference on both main effects (position along the tract; diagnostic group) and their interactions is made using permutation testing strategies that do not require distributional assumptions. Also, to investigate the question of homogeneity we implement a permutation testing strategy, which adapts a "block bootstrapping" approach from time series analysis to the functional data setting.

The Impact of COVID-19 Lockdown on Daily Activities, Cognitions, and Stress in a Lonely and Distressed Population: Temporal Dynamic Network Analysis.

BACKGROUND: The COVID-19 pandemic and its associated lockdown measures impacted mental health worldwide. However, the temporal dynamics of causal factors that modulate mental health during lockdown are not well understood. OBJECTIVE: We aimed to understand how a COVID-19 lockdown changes the temporal dynamics of loneliness and other factors affecting mental health. This is the first study that compares network characteristics between lockdown stages to prioritize mental health intervention targets. METHODS: We combined ecological momentary assessments with wrist-worn motion tracking to investigate the mechanism and changes in network centrality of symptoms and behaviors before and during lockdown. A total of 258 participants who reported at least mild loneliness and distress were assessed 8 times a day for 7 consecutive days over a 213-day period from August 8, 2020, through March 9, 2021, in Germany, covering a "no-lockdown" and a "lockdown" stage. COVID-19-related worry, information-seeking, perceived restriction, and loneliness were assessed by digital visual analog scales ranging from 0 to 100. Social activity was assessed on a 7-point Likert scale, while physical activity was recorded from wrist-worn actigraphy devices. RESULTS: We built a multilevel vector autoregressive model to estimate dynamic networks. To compare network characteristics between a no-lockdown stage and a lockdown stage, we performed permutation tests. During lockdown, loneliness had the highest impact within the network, as indicated by its centrality index (ie, an index to identify variables that have a strong influence on the other variables). Moreover, during lockdown, the centrality of loneliness significantly increased. Physical activity contributed to a decrease in loneliness amid the lockdown stage. CONCLUSIONS: The COVID-19 lockdown increased the central role of loneliness in triggering stress-related behaviors and cognition. Our study indicates that loneliness should be prioritized in mental health interventions during lockdown. Moreover, physical activity can serve as a buffer for loneliness amid social restrictions.

Optimized permutation testing for information theoretic measures of multi-gene interactions.

BACKGROUND: Permutation testing is often considered the "gold standard" for multi-test significance analysis, as it is an exact test requiring few assumptions about the distribution being computed. However, it can be computationally very expensive, particularly in its naive form in which the full analysis pipeline is re-run after permuting the phenotype labels. This can become intractable in multi-locus genome-wide association studies (GWAS), in which the number of potential interactions to be tested is combinatorially large. RESULTS: In this paper, we develop an approach for permutation testing in multi-locus GWAS, specifically focusing on SNP-SNP-phenotype interactions using multivariable measures that can be computed from frequency count tables, such as those based in Information Theory. We find that the computational bottleneck in this process is the construction of the count tables themselves, and that this step can be eliminated at each iteration of the permutation testing by transforming the count tables directly. This leads to a speed-up by a factor of over 103 for a typical permutation test compared to the naive approach. Additionally, this approach is insensitive to the number of samples making it suitable for datasets with large number of samples. CONCLUSIONS: The proliferation of large-scale datasets with genotype data for hundreds of thousands of individuals enables new and more powerful approaches for the detection of multi-locus genotype-phenotype interactions. Our approach significantly improves the computational tractability of permutation testing for these studies. Moreover, our approach is insensitive to the large number of samples in these modern datasets. The code for performing these computations and replicating the figures in this paper is freely available at https://github.com/kunert/permute-counts .

A simple permutation-based test of intermodal correspondence.

Many key findings in neuroimaging studies involve similarities between brain maps, but statistical methods used to measure these findings have varied. Current state-of-the-art methods involve comparing observed group-level brain maps (after averaging intensities at each image location across multiple subjects) against spatial null models of these group-level maps. However, these methods typically make strong and potentially unrealistic statistical assumptions, such as covariance stationarity. To address these issues, in this article we propose using subject-level data and a classical permutation testing framework to test and assess similarities between brain maps. Our method is comparable to traditional permutation tests in that it involves randomly permuting subjects to generate a null distribution of intermodal correspondence statistics, which we compare to an observed statistic to estimate a p-value. We apply and compare our method in simulated and real neuroimaging data from the Philadelphia Neurodevelopmental Cohort. We show that our method performs well for detecting relationships between modalities known to be strongly related (cortical thickness and sulcal depth), and it is conservative when an association would not be expected (cortical thickness and activation on the n-back working memory task). Notably, our method is the most flexible and reliable for localizing intermodal relationships within subregions of the brain and allows for generalizable statistical inference.

Inference in functional mixed regression models with applications to Positron Emission Tomography imaging data.

In a function-on-scalar regression framework, we present some modeling strategies for functional mixed models and also some approaches for making inference about various aspects of the fixed effects. This is presented in the context of modeling positron emission tomography (PET) data in order to explore the density of various proteins of interest throughout the human brain. For this application, information about the density of the target protein in a given brain region is encapsulated in the impulse response function (IRF) of the region. Previous work on nonparametric estimation of the IRF is limited in that it is only able to model a single brain region at a time. We propose an extension, based on principles of functional data analysis, that will allow modeling of multiple brain regions simultaneously. Applicable more broadly to functional mixed regression modeling, we discuss two general approaches for permutation testing and describe valid strategies for identifying exchangeable units within the model and building corresponding permutation tests. We illustrate our methods with an application to PET data and explore the effects of depression and sex on the IRF.

Network effects in influenza spread: The impact of mobility and socio-economic factors.

This paper introduces new methods of modeling and analyzing social networks that emerge in the context of disease spread. Four methods of constructing informative networks are presented, two of which use. static data and two use temporal data, namely individual citizen mobility observations taken over an extensive period of time. We show how the built networks can be analyzed, and how the numerical results can be interpreted, using network permutation-based surprise analysis. In doing so, we explain the relationship of surprise analysis with conventional network hypothesis testing and Quadratic Assignment Procedure regression. Surprise analysis is more comprehensive, and can be without limitation performed with any form(s) of network subgraphs, including those with multiple nodal attributes, weighted links, and temporal features. To illustrate our methodological work in application, we put them to use for interpreting networks constructed from the data collected over one year in an observational study in Buffalo and Erie counties in New York state during the 2016-2017 influenza season. Even with the limitations in the data size, our methods are able to reveal the global (city- and season-wide) patterns in the spread of influenza, taking into account population mobility and socio-economic factors.

Permutation testing of Fourier shell correlation for resolution estimation of cryo-EM maps.

Fourier shell correlation (FSC) has become a standard quantity for resolution estimation in electron cryo-microscopy. However, the resolution determination step is still subjective and not fully automated as it involves a series of map interventions before FSC computation and includes the selection of a common threshold. Here, we apply the statistical methods of permutation testing and false discovery rate (FDR) control to the resolution-dependent correlation measure. The approach allows fully automated and mask-free resolution determination based on statistical thresholding of FSC curves. We demonstrate the applicability for global, local and directional resolution estimation and show that the developed criterion termed FDR-FSC gives realistic resolution estimates based on statistical significance while eliminating the need of any map manipulations. The algorithms are implemented in a user-friendly GUI based software tool termed SPoC (https://github.com/MaximilianBeckers/SPOC).

Cerebral cortical thickness and surface area in adolescent anorexia nervosa: Separate and joint analyses with a permutation-based nonparametric method.

OBJECTIVE: Reduction in cerebral volume is often found in underweight patients with anorexia nervosa (AN), but few studies have investigated other morphological measures. Cortical thickness (CTh) and surface area (CSA), often used to produce the measure of cortical volume, are developmentally distinct measures that may be differentially affected in AN, particularly in the developing brain. In the present study, we investigated CTh and CSA both separately and jointly to gain further insight into structural alterations in adolescent AN patients. METHOD: Thirty female AN inpatients 12-18 years of age, and 27 age-matched healthy controls (HC) underwent structural magnetic resonance imaging. Group differences in CTh and CSA were investigated separately and jointly with a permutation-based nonparametric combination method (NPC) which may be more sensitive in detecting group differences compared to traditional volumetric methods. RESULTS: Results showed significant reduction in in both CTh and CSA in several cortical regions in AN compared to HC and the reduction was related to BMI. Different results for the two morphological measures were found in a small number of cortical regions. The joint NPC analyses showed significant group differences across most of the cortical mantle. DISCUSSION: Results from this study give novel insight to areal reduction in adolescent AN patients and indicate that both CTh and CSA reduction is related to BMI. The study is the first to use the NPC method to reveal large structural alterations covering most of the brain in adolescent AN.

Stable between-subject statistical inference from unstable within-subject functional connectivity estimates.

Spatial or temporal aspects of neural organization are known to be important indices of how cognition is organized. However, measurements and estimations are often noisy and many of the algorithms used are probabilistic, which in combination have been argued to limit studies exploring the neural basis of specific aspects of cognition. Focusing on static and dynamic functional connectivity estimations, we propose to leverage this variability to improve statistical efficiency in relating these estimations to behavior. To achieve this goal, we use a procedure based on permutation testing that provides a way of combining the results from many individual tests that refer to the same hypothesis. This is needed when testing a measure whose value is obtained from a noisy process, which can be repeated multiple times, referred to as replications. Focusing on functional connectivity, this noisy process can be: (a) computational, for example, when using an approximate inference algorithm for which different runs can produce different results or (b) observational, if we have the capacity to acquire data multiple times, and the different acquired data sets can be considered noisy examples of some underlying truth. In both cases, we are not interested in the individual replications but on the unobserved process generating each replication. In this note, we show how results can be combined instead of choosing just one of the estimated models. Using both simulations and real data, we show the benefits of this approach in practice.

MIDAS: Regionally linear multivariate discriminative statistical mapping.

Statistical parametric maps formed via voxel-wise mass-univariate tests, such as the general linear model, are commonly used to test hypotheses about regionally specific effects in neuroimaging cross-sectional studies where each subject is represented by a single image. Despite being informative, these techniques remain limited as they ignore multivariate relationships in the data. Most importantly, the commonly employed local Gaussian smoothing, which is important for accounting for registration errors and making the data follow Gaussian distributions, is usually chosen in an ad hoc fashion. Thus, it is often suboptimal for the task of detecting group differences and correlations with non-imaging variables. Information mapping techniques, such as searchlight, which use pattern classifiers to exploit multivariate information and obtain more powerful statistical maps, have become increasingly popular in recent years. However, existing methods may lead to important interpretation errors in practice (i.e., misidentifying a cluster as informative, or failing to detect truly informative voxels), while often being computationally expensive. To address these issues, we introduce a novel efficient multivariate statistical framework for cross-sectional studies, termed MIDAS, seeking highly sensitive and specific voxel-wise brain maps, while leveraging the power of regional discriminant analysis. In MIDAS, locally linear discriminative learning is applied to estimate the pattern that best discriminates between two groups, or predicts a variable of interest. This pattern is equivalent to local filtering by an optimal kernel whose coefficients are the weights of the linear discriminant. By composing information from all neighborhoods that contain a given voxel, MIDAS produces a statistic that collectively reflects the contribution of the voxel to the regional classifiers as well as the discriminative power of the classifiers. Critically, MIDAS efficiently assesses the statistical significance of the derived statistic by analytically approximating its null distribution without the need for computationally expensive permutation tests. The proposed framework was extensively validated using simulated atrophy in structural magnetic resonance imaging (MRI) and further tested using data from a task-based functional MRI study as well as a structural MRI study of cognitive performance. The performance of the proposed framework was evaluated against standard voxel-wise general linear models and other information mapping methods. The experimental results showed that MIDAS achieves relatively higher sensitivity and specificity in detecting group differences. Together, our results demonstrate the potential of the proposed approach to efficiently map effects of interest in both structural and functional data.

Heterogeneous public and local knowledge provides a qualitative indicator of coastal use by marine recreational fishers.

Marine recreational fishing (MRF) benefits individuals and economies, but can also impact fish stocks and associated ecosystems. Fish are an important resource providing direct economic benefit through commercial and recreational exploitation, and more esoteric ecosystem services. It is important to consider recreational fishing in marine spatial planning, but spatial information on coastal utilisation for MRF is frequently lacking. Public sources of local knowledge were reviewed and the frequency of unique references to sites extracted. Sites were georeferenced using a gazetteer compiled from the Ordnance Survey and United Kingdom Hydrographic Office named sea features gazetteer and local knowledge sources. Recreational fishing site densities were calculated across 2700 km of coastline and this proxy indicator of coastal utilisation validated against two independent surveys using permutative Monte Carlo sampling to control for sparse and non-independent data. Site density had fair agreement with independent surveys, but standardization by shore length reduced this agreement. Applying a 3 by 3 box filter convolution to the spatial layers improved the agreement between local knowledge derived predictions of activity and those of directed surveys, and permutation testing showed that agreement did not arise as a result of the convolution itself. High and low activity areas were more accurately predicted than areas of intermediate activity. Site density derived from heterogeneous participant and local knowledge can produce qualitative predictions of where recreational fishers fish, and applying a convolution can improve the predictive power of data so derived. However, this approach will be subject to unquantifiable bias and may fail to identify areas highly valued by marine recreational fishers. Thus it should be used in conjunction with other information in decision making and may be best suited to inform the early stage sampling design of on-site surveys or to complement other data sets in mapping areas of importance to recreational fishers.

PredPsych: A toolbox for predictive machine learning-based approach in experimental psychology research.

Recent years have seen an increased interest in machine learning-based predictive methods for analyzing quantitative behavioral data in experimental psychology. While these methods can achieve relatively greater sensitivity compared to conventional univariate techniques, they still lack an established and accessible implementation. The aim of current work was to build an open-source R toolbox - "PredPsych" - that could make these methods readily available to all psychologists. PredPsych is a user-friendly, R toolbox based on machine-learning predictive algorithms. In this paper, we present the framework of PredPsych via the analysis of a recently published multiple-subject motion capture dataset. In addition, we discuss examples of possible research questions that can be addressed with the machine-learning algorithms implemented in PredPsych and cannot be easily addressed with univariate statistical analysis. We anticipate that PredPsych will be of use to researchers with limited programming experience not only in the field of psychology, but also in that of clinical neuroscience, enabling computational assessment of putative bio-behavioral markers for both prognosis and diagnosis.

Statistical Permutation-based Artery Mapping (SPAM): a novel approach to evaluate imaging signals in the vessel wall.

BACKGROUND: Cardiovascular diseases are the leading cause of death worldwide. A prominent cause of cardiovascular events is atherosclerosis, a chronic inflammation of the arterial wall that leads to the formation of so called atherosclerotic plaques. There is a strong clinical need to develop new, non-invasive vascular imaging techniques in order to identify high-risk plaques, which might escape detection using conventional methods based on the assessment of the luminal narrowing. In this context, molecular imaging strategies based on fluorescent tracers and fluorescence reflectance imaging (FRI) seem well suited to assess molecular and cellular activity. However, such an analysis demands a precise and standardized analysis method, which is orientated on reproducible anatomical landmarks, ensuring to compare equivalent regions across different subjects. METHODS: We propose a novel method, Statistical Permutation-based Artery Mapping (SPAM). Our approach is especially useful for the understanding of complex and heterogeneous regional processes during the course of atherosclerosis. Our method involves three steps, which are (I) standardisation with an additional intensity normalization, (II) permutation testing, and (III) cluster-enhancement. Although permutation testing and cluster enhancement are already well-established in functional magnetic resonance imaging, to the best of our knowledge these strategies have so far not been applied in cardiovascular molecular imaging. RESULTS: We tested our method using FRI images of murine aortic vessels in order to find recurring patterns in atherosclerotic plaques across multiple subjects. We demonstrate that our pixel-wise and cluster-enhanced testing approach is feasible and useful to analyse tracer distributions in FRI data sets of aortic vessels. CONCLUSIONS: We expect our method to be a useful tool within the field of molecular imaging of atherosclerotic plaques since cluster-enhanced permutation testing is a powerful approach for finding significant differences of tracer distributions in inflamed atherosclerotic vessels.

Combining p-values in replicated single-case experiments with multivariate outcome.

Interest in combining probabilities has a long history in the global statistical community. The first steps in this direction were taken by Ronald Fisher, who introduced the idea of combining p-values of independent tests to provide a global decision rule when multiple aspects of a given problem were of interest. An interesting approach to this idea of combining p-values is the one based on permutation theory. The methods belonging to this particular approach exploit the permutation distributions of the tests to be combined, and use a simple function to combine probabilities. Combining p-values finds a very interesting application in the analysis of replicated single-case experiments. In this field the focus, while comparing different treatments effects, is more articulated than when just looking at the means of the different populations. Moreover, it is often of interest to combine the results obtained on the single patients in order to get more global information about the phenomenon under study. This paper gives an overview of how the concept of combining p-values was conceived, and how it can be easily handled via permutation techniques. Finally, the method of combining p-values is applied to a simulated replicated single-case experiment, and a numerical illustration is presented.

Functional Analysis of G6PD Variants Associated With Low G6PD Activity in the All of Us Research Program.

Glucose-6-phosphate dehydrogenase (G6PD) protects red blood cells against oxidative damage through regeneration of NADPH. Individuals with G6PD polymorphisms (variants) that produce an impaired G6PD enzyme are usually asymptomatic, but at risk of hemolytic anemia from oxidative stressors, including certain drugs and foods. Prevention of G6PD deficiency-related hemolytic anemia is achievable through G6PD genetic testing or whole-genome sequencing (WGS) to identify affected individuals who should avoid hemolytic triggers. However, accurately predicting the clinical consequence of G6PD variants is limited by over 800 G6PD variants which remain of uncertain significance. There also remains significant variability in which deficiency-causing variants are included in pharmacogenomic testing arrays across institutions: many panels only include c.202G>A, even though dozens of other variants can also cause G6PD deficiency. Here, we seek to improve G6PD genotype interpretation using data available in the All of Us Research Program and using a yeast functional assay. We confirm that G6PD coding variants are the main contributor to decreased G6PD activity, and that 13% of individuals in the All of Us data with deficiency-causing variants would be missed if only the c.202G>A variant were tested for. We expand clinical interpretation for G6PD variants of uncertain significance; reporting that c.595A>G, known as G6PD Dagua or G6PD Açores, and the newly identified variant c.430C>G, reduce activity sufficiently to lead to G6PD deficiency. We also provide evidence that five missense variants of uncertain significance are unlikely to lead to G6PD deficiency, since they were seen in hemi- or homozygous individuals without a reduction in G6PD activity. We also applied the new WHO guidelines and were able to classify two synonymous variants as WHO class C. We anticipate these results will improve the accuracy, and prompt increased use, of G6PD genetic tests through a more complete clinical interpretation of G6PD variants. As the All of Us data increases from 245,000 to 1 million participants, and additional functional assays are carried out, we expect this research to serve as a template to enable complete characterization of G6PD deficiency genotypes. With an increased number of interpreted variants, genetic testing of G6PD will be more informative for preemptively identifying individuals at risk for drug- or food-induced hemolytic anemia.

Exact inference around ordinal measures of association is often not exact.

In this paper, we build upon the work of DiCiccio and Romano (2017) by extending their permutation test approach, based on the Pearson correlation coefficient in the continuous case, to ordinal measures of association. We investigate commonly used ordinal measures such as the Spearman correlation, Kendall's tau-b, and gamma, which are widely implemented in commercial and open-source software packages for exact testing routines based on generalized hypergeometric probabilities. Similar to DiCiccio and Romano's method, we apply studentization to correct the test statistic, which yields asymptotically valid inference for testing no ordinal association. We present a comprehensive theoretical framework for our approach, followed by a simulation study. Furthermore, we use toy examples to highlight the differences between the exact tests and the asymptotically valid tests. Our findings align with those of DiCiccio and Romano, indicating that exact permutation tests based on ordinal measures of association are often not exact, whereas the asymptotically correct tests perform well for moderate to large sample sizes.

Bridging the Gap between Deep Learning and Hypothesis-Driven Analysis via Permutation Testing.

A fundamental approach in neuroscience research is to test hypotheses based on neuropsychological and behavioral measures, i.e., whether certain factors (e.g., related to life events) are associated with an outcome (e.g., depression). In recent years, deep learning has become a potential alternative approach for conducting such analyses by predicting an outcome from a collection of factors and identifying the most "informative" ones driving the prediction. However, this approach has had limited impact as its findings are not linked to statistical significance of factors supporting hypotheses. In this article, we proposed a flexible and scalable approach based on the concept of permutation testing that integrates hypothesis testing into the data-driven deep learning analysis. We apply our approach to the yearly self-reported assessments of 621 adolescent participants of the National Consortium of Alcohol and Neurodevelopment in Adolescence (NCANDA) to predict negative valence, a symptom of major depressive disorder according to the NIMH Research Domain Criteria (RDoC). Our method successfully identifies categories of risk factors that further explain the symptom.

Fecal microbiota and metabolomics revealed the effect of long-term consumption of gallic acid on canine lipid metabolism and gut health.

Gallic acid (GA) is a natural polyphenolic compound with many health benefits. To assess the potential risk of long-term consumption of GA to gut health, healthy dogs were fed a basal diet supplemented with GA (0%, 0.02%, 0.04%, and 0.08%) for 45 d, and fecal microbiota and metabolomics were evaluated. This study demonstrated that GA supplementation regulated serum lipid metabolism by reducing serum triglyceride, fat digestibility, and Bacteroidetes/Firmicutes ratio. In addition, the relative abundance of Parasutterella was significantly lower, and the SCFAs-producing bacteria were increased along with fecal acetate and total SCFAs contents accumulation in the 0.08% GA group. Metabolomics data further elucidated that 0.08% GA significantly affected carbohydrate metabolism by downregulating succinic acid in fece, thereby alleviating inflammation and oxidative stress. Overall, this study confirmed the beneficial effects of long-term consumption of GA on lipid metabolism and gut health, and the optimal level of GA supplementation was 0.08%.

Use of ultra high performance liquid chromatography with high resolution mass spectrometry to analyze urinary metabolome alterations following acute kidney injury in post-cardiac surgery patients.

BACKGROUND: Cardiac surgery-associated acute kidney injury (AKI) can increase the mortality and morbidity, and the incidence of chronic kidney disease, in critically ill survivors. The purpose of this research was to investigate possible links between urinary metabolic changes and cardiac surgery-associated AKI. METHODS: Using ultra-high-performance liquid chromatography coupled with Q-Exactive Orbitrap mass spectrometry, non-targeted metabolomics was performed on urinary samples collected from groups of patients with cardiac surgery-associated AKI at different time points, including Before_AKI (uninjured kidney), AKI_Day1 (injured kidney) and AKI_Day14 (recovered kidney) groups. The data among the three groups were analyzed by combining multivariate and univariate statistical methods, and urine metabolites related to AKI in patients after cardiac surgery were screened. Altered metabolic pathways associated with cardiac surgery-induced AKI were identified by examining the Kyoto Encyclopedia of Genes and Genomes database. RESULTS: The secreted urinary metabolome of the injured kidney can be well separated from the urine metabolomes of uninjured or recovered patients using multivariate and univariate statistical analyses. However, urine samples from the AKI_Day14 and Before_AKI groups cannot be distinguished using either of the two statistical analyses. Nearly 4000 urinary metabolites were identified through bioinformatics methods at Annotation Levels 1-4. Several of these differential metabolites may also perform essential biological functions. Differential analysis of the urinary metabolome among groups was also performed to provide potential prognostic indicators and changes in signalling pathways. Compared with the uninjured kidney group, the patients with cardiac surgery-associated AKI displayed dramatic changes in renal metabolism, including sulphur metabolism and amino acid metabolism. CONCLUSIONS: Urinary metabolite disorder was observed in patients with cardiac surgery-associated AKI due to ischaemia and medical treatment, and the recovered patients' kidneys were able to return to normal. This work provides data on urine metabolite markers and essential resources for further research on AKI.