A generalized covariate-adjusted top-scoring pair algorithm with applications to diabetic kidney disease stage classification in the Chronic Renal Insufficiency Cohort (CRIC) Study

BACKGROUND

The growing amount of high dimensional biomolecular data has spawned new statistical and computational models for risk prediction and disease classification. Yet, many of these methods do not yield biologically interpretable models, despite offering high classification accuracy. An exception, the top-scoring pair (TSP) algorithm derives parameter-free, biologically interpretable single pair decision rules that are accurate and robust in disease classification. However, standard TSP methods do not accommodate covariates that could heavily influence feature selection for the top-scoring pair. Herein, we propose a covariate-adjusted TSP method, which uses residuals from a regression of features on the covariates for identifying top scoring pairs. We conduct simulations and a data application to investigate our method, and compare it to existing classifiers, LASSO and random forests.

RESULTS

Our simulations found that features that were highly correlated with clinical variables had high likelihood of being selected as top scoring pairs in the standard TSP setting. However, through residualization, our covariate-adjusted TSP was able to identify new top scoring pairs, that were largely uncorrelated with clinical variables. In the data application, using patients with diabetes (n = 977) selected for metabolomic profiling in the Chronic Renal Insufficiency Cohort (CRIC) study, the standard TSP algorithm identified (valine-betaine, dimethyl-arg) as the top-scoring metabolite pair for classifying diabetic kidney disease (DKD) severity, whereas the covariate-adjusted TSP method identified the pair (pipazethate, octaethylene glycol) as top-scoring. Valine-betaine and dimethyl-arg had, respectively, ≥ 0.4 absolute correlation with urine albumin and serum creatinine, known prognosticators of DKD. Thus without covariate-adjustment the top-scoring pair largely reflected known markers of disease severity, whereas covariate-adjusted TSP uncovered features liberated from confounding, and identified independent prognostic markers of DKD severity. Furthermore, TSP-based methods achieved competitive classification accuracy in DKD to LASSO and random forests, while providing more parsimonious models.

CONCLUSIONS

We extended TSP-based methods to account for covariates, via a simple, easy to implement residualizing process. Our covariate-adjusted TSP method identified metabolite features, uncorrelated from clinical covariates, that discriminate DKD severity stage based on the relative ordering between two features, and thus provide insights into future studies on the order reversals in early vs advanced disease states.

Using transcriptomics to predict and visualize disease status in bighorn sheep (Ovis canadensis)

Increasing risk of pathogen spillover coupled with overall declines in wildlife population abundance in the Anthropocene make infectious disease a relevant concern for species conservation worldwide. While emerging molecular tools could improve our diagnostic capabilities and give insight into mechanisms underlying wildlife disease risk, they have rarely been applied in practice. Here, employing a previously reported gene transcription panel of common immune markers to track physiological changes, we present a detailed analysis over the course of both acute and chronic infection in one wildlife species where disease plays a critical role in conservation, bighorn sheep (Ovis canadensis). Differential gene transcription patterns distinguished between infection statuses over the course of acute infection and differential correlation (DC) analyses identified clear changes in gene co-transcription patterns over the early stages of infection, with transcription of four genes-TGFb, AHR, IL1b and MX1-continuing to increase even as transcription of other immune-associated genes waned. In a separate analysis, we considered the capacity of the same gene transcription panel to aid in differentiating between chronically infected animals and animals in other disease states outside of acute disease events (an immediate priority for wildlife management in this system). We found that this transcription panel was capable of accurately identifying chronically infected animals in the test dataset, though additional data will be required to determine how far this ability extends. Taken together, our results showcase the successful proof of concept and breadth of potential utilities that gene transcription might provide to wildlife disease management, from direct insight into mechanisms associated with differential disease response to improved diagnostic capacity in the field.

Advanced machine learning methods in psychiatry: an introduction

Mental health questions can be tackled through machine learning (ML) techniques. Apart from the two ML methods we introduced in our previous paper, we discuss two more advanced ML approaches in this paper: support vector machines and artificial neural networks. To illustrate how these ML methods have been employed in mental health, recent research applications in psychiatry were reported.

Covariate adjusted classification trees

In studies that compare several diagnostic groups, subjects can be measured on certain features and classification trees can be used to identify which of them best characterize the differences among groups. However, subjects may also be measured on additional covariates whose ability to characterize group differences is not meaningful or of interest, but may still have an impact on the examined features. Therefore, it is important to adjust for the effects of covariates on these features. We present a new semi-parametric approach to adjust for covariate effects when constructing classification trees based on the features of interest that is readily implementable. An application is given for postmortem brain tissue data to compare the neurobiological characteristics of subjects with schizophrenia to those of normal controls. We also evaluate the performance of our approach using a simulation study.

Adjusting for matching and covariates in linear discriminant analysis

In studies that compare several diagnostic or treatment groups, subjects may not only be measured on a certain set of feature variables, but also be matched on a number of demographic characteristics and measured on additional covariates. Linear discriminant analysis (LDA) is sometimes used to identify which feature variables best discriminate among groups, while accounting for the dependencies among the feature variables. We present a new approach to LDA for multivariate normal data that accounts for the subject matching used in a particular study design, as well as covariates not used in the matching. Applications are given for post-mortem tissue data with the aim of comparing neurobiological characteristics of subjects with schizophrenia with those of normal controls, and for a post-mortem tissue primate study comparing brain biomarker measurements across three treatment groups. We also investigate the performance of our approach using a simulation study.

Generalized covariance-adjusted canonical correlation analysis with application to psychiatry

The lack of control over covariates in practice motivates the need for their adjustment when measuring the degree of association between two sets of variables, for which canonical correlation is traditionally used. In most studies however, there is also a lack of control over the attributes of responses for the sets of variables of interest. In particular, a portion of the response variable may be continuous and the other discrete. For such settings, the traditional partial canonical correlation approach is restrictive, since a covariate-adjustment for a set of continuous variables is assumed. By ignoring the assumption of continuous variates and proceeding with a partial canonical correlation analysis in the presence of continuous and discrete variates, results in canonical correlation estimates that are not consistent. In this paper we generalize the traditional partial canonical correlation approach to covariate-adjustment by allowing the response variables to contain continuous, as well as discrete, variates. The methodology is illustrated with a psychiatric application for examining which sleep variables relate to which depressive symptoms, as measured by commonly used constructs that presents with both continuous and discrete outcomes.