An introspective comparison of random forest-based classifiers for the analysis of cluster-correlated data by way of RF++

Fixed Effects or Mixed Effects Classifiers? Evidence From Simulated and Archival Data

This study seeks to compare fixed and mixed effects models for the purposes of predictive classification in the presence of multilevel data. The first part of the study utilizes a Monte Carlo simulation to compare fixed and mixed effects logistic regression and random forests. An applied examination of the prediction of student retention in the public-use U.S. PISA data set was considered to verify the simulation findings. Results of this study indicate fixed effects models performed comparably with mixed effects models across both the simulation and PISA examinations. Results broadly suggest that researchers should be cognizant of the type of predictors and data structure being used, as these factors carried more weight than did the model type.

Discordant and Converting Receptor Expressions in Brain Metastases from Breast Cancer: MRI-Based Non-Invasive Receptor Status Tracking

Discordance and conversion of receptor expressions in metastatic lesions and primary tumors is often observed in patients with brain metastases from breast cancer. Therefore, personalized therapy requires continuous monitoring of receptor expressions and dynamic adaptation of applied targeted treatment options. Radiological in vivo techniques may allow receptor status tracking at high frequencies at low risk and cost. The present study aims to investigate the potential of receptor status prediction through machine-learning-based analysis of radiomic MR image features. The analysis is based on 412 brain metastases samples from 106 patients acquired between 09/2007 and 09/2021. Inclusion criteria were as follows: diagnosed cerebral metastases from breast cancer; histopathology reports on progesterone (PR), estrogen (ER), and human epidermal growth factor 2 (HER2) receptor status; and availability of MR imaging data. In total, 3367 quantitative features of T1 contrast-enhanced, T1 non-enhanced, and FLAIR images and corresponding patient age were evaluated utilizing random forest algorithms. Feature importance was assessed using Gini impurity measures. Predictive performance was tested using 10 permuted 5-fold cross-validation sets employing the 30 most important features of each training set. Receiver operating characteristic areas under the curves of the validation sets were 0.82 (95% confidence interval [0.78; 0.85]) for ER+, 0.73 [0.69; 0.77] for PR+, and 0.74 [0.70; 0.78] for HER2+. Observations indicate that MR image features employed in a machine learning classifier could provide high discriminatory accuracy in predicting the receptor status of brain metastases from breast cancer.

A review on longitudinal data analysis with random forest

In longitudinal studies variables are measured repeatedly over time, leading to clustered and correlated observations. If the goal of the study is to develop prediction models, machine learning approaches such as the powerful random forest (RF) are often promising alternatives to standard statistical methods, especially in the context of high-dimensional data. In this paper, we review extensions of the standard RF method for the purpose of longitudinal data analysis. Extension methods are categorized according to the data structures for which they are designed. We consider both univariate and multivariate response longitudinal data and further categorize the repeated measurements according to whether the time effect is relevant. Even though most extensions are proposed for low-dimensional data, some can be applied to high-dimensional data. Information of available software implementations of the reviewed extensions is also given. We conclude with discussions on the limitations of our review and some future research directions.

Metabolomics and random forests in patients with complex congenital heart disease

Introduction

It is increasingly common to simultaneously determine a large number of metabolites in order to assess the metabolic state of, or clarify biochemical pathways in, an organism (“metabolomics”). This approach is increasingly used in the investigation of the development of heart failure. Recently, the first reports with respect to a metabolomic approach for the assessment of patients with complex congenital heart disease have been published. Classical statistical analysis of such data is challenging.

Objective

This study aims to present an alternative to classical statistics with respect to identifying relevant metabolites in a classification task and numerically estimating their relative impact.

Methods

Data from two metabolomic studies on 20 patients with complex congenital heart disease and Fontan circulation and 20 controls were reanalysed using random forest (RF) methodology. Results were compared to those of classical statistics.

Results

RF analysis required no elaborate data pre-processing. The ranking of the variables with respect to classification impact (subject diseased, or not) was remarkably similar irrespective of the evaluation method used, leading to identical clinical interpretation.

Conclusion

In metabolomic classification in adult patients with complex congenital heart disease, RF analysis as a one-step method delivers the most adequate results with minimum effort. RF may serve as an adjunct to traditional statistics also in this small but crucial-to-monitor patient group.

Segmentation of PMSE Data Using Random Forests

EISCAT VHF radar data are used for observing, monitoring, and understanding Earth’s upper atmosphere. This paper presents an approach to segment Polar Mesospheric Summer Echoes (PMSE) from datasets obtained from EISCAT VHF radar data. The data consist of 30 observations days, corresponding to 56,250 data samples. We manually labeled the data into three different categories: PMSE, Ionospheric background, and Background noise. For segmentation, we employed random forests on a set of simple features. These features include: altitude derivative, time derivative, mean, median, standard deviation, minimum, and maximum values corresponding to neighborhood sizes ranging from 3 by 3 to 11 by 11 pixels. Next, in order to reduce the model bias and variance, we employed a method that decreases the weight applied to pixel labels with large uncertainty. Our results indicate that, first, it is possible to segment PMSE from the data using random forests. Second, the weighted-down labels technique improves the performance of the random forests method.

Evaluation of magnetic resonance imaging-based radiomics characteristics for differentiation of benign and malignant peripheral nerve sheath tumors in neurofibromatosis type 1

BACKGROUND

Patients with neurofibromatosis type 1 (NF1) develop benign (BPNST), premalignant atypical (ANF), and malignant (MPNST) peripheral nerve sheath tumors. Radiological differentiation of these entities is challenging. Therefore, we aimed to evaluate the value of a magnetic resonance imaging (MRI)-based radiomics machine-learning (ML) classifier for differentiation of these three entities of internal peripheral nerve sheath tumors in NF1 patients.

METHODS

MRI was performed at 3T in 36 NF1 patients (20 male; age: 31 ± 11 years). Segmentation of 117 BPNSTs, 17 MPNSTs, and 8 ANFs was manually performed using T2w spectral attenuated inversion recovery sequences. One hundred seven features per lesion were extracted using PyRadiomics and applied for BPNST versus MPNST differentiation. A 5-feature radiomics signature was defined based on the most important features and tested for signature-based BPNST versus MPNST classification (random forest [RF] classification, leave-one-patient-out evaluation). In a second step, signature feature expressions for BPNSTs, ANFs, and MPNSTs were evaluated for radiomics-based classification for these three entities.

RESULTS

The mean area under the receiver operator characteristic curve (AUC) for the radiomics-based BPNST versus MPNST differentiation was 0.94, corresponding to correct classification of on average 16/17 MPNSTs and 114/117 BPNSTs (sensitivity: 94%, specificity: 97%). Exploratory analysis with the eight ANFs revealed intermediate radiomic feature characteristics in-between BPNST and MPNST tumor feature expression.

CONCLUSION

In this proof-of-principle study, ML using MRI-based radiomics characteristics allows sensitive and specific differentiation of BPNSTs and MPNSTs in NF1 patients. Feature expression of premalignant atypical tumors was distributed in-between benign and malignant tumor feature expressions, which illustrates biological plausibility of the considered radiomics characteristics.

A Machine Learning Approach for Investigating Delirium as a Multifactorial Syndrome

Delirium is a psycho-organic syndrome common in hospitalized patients, especially the elderly, and is associated with poor clinical outcomes. This study aims to identify the predictors that are mostly associated with the risk of delirium episodes using a machine learning technique (MLT). A random forest (RF) algorithm was used to evaluate the association between the subject’s characteristics and the 4AT (the 4 A’s test) score screening tool for delirium. RF algorithm was implemented using information based on demographic characteristics, comorbidities, drugs and procedures. Of the 78 patients enrolled in the study, 49 (63%) were at risk for delirium, 32 (41%) had at least one episode of delirium during the hospitalization (38% in orthopedics and 31% both in internal medicine and in the geriatric ward). The model explained 75.8% of the variability of the 4AT score with a root mean squared error of 3.29. Higher age, the presence of dementia, physical restraint, diabetes and a lower degree are the variables associated with an increase of the 4AT score. Random forest is a valid method for investigating the patients’ characteristics associated with delirium onset also in small case-series. The use of this model may allow for early detection of delirium onset to plan the proper adjustment in healthcare assistance.

Control of n-Butanol Induced Lipidome Adaptations in E. coli

The versatile compound n-butanol is one of the most promising biofuels for use in existing internal combustion engines, contributing to a smooth transition towards a clean energy society. Furthermore, n-butanol is a valuable resource to produce more complex molecules such as bioplastics. Microbial production of n-butanol from waste materials is hampered by the biotoxicity of n-butanol as it interferes with the proper functioning of lipid membranes. In this study we perform a large-scale investigation of the complete lipid-related enzyme machinery and its response to exposure to a sublethal concentration of n-butanol. We profiled, in triplicate, the growth characteristics and phospholipidomes of 116 different genetic constructs of E. coli, both in the presence and absence of 0.5% n-butanol (v/v). This led to the identification of 230 lipid species and subsequently to the reconstruction of the network of metabolites, enzymes and lipid properties driving the homeostasis of the E. coli lipidome. We were able to identify key lipids and biochemical pathways leading to altered n-butanol tolerance. The data led to new conceptual insights into the bacterial lipid metabolism which are discussed.

Generalized linear mixed-model (GLMM) trees: A flexible decision-tree method for multilevel and longitudinal data

Objective: Decision-tree methods are machine-learning methods which provide results that are relatively easy to interpret and apply by human decision makers. The resulting decision trees show how baseline patient characteristics can be combined to predict treatment outcomes for individual patients, for example. This paper introduces GLMM trees, a decision-tree method for multilevel and longitudinal data. Method: To illustrate, we apply GLMM trees to a dataset of 3,256 young people (mean age 11.33, 48% girls) receiving treatment at one of several mental-health service providers in the UK. Two treatment outcomes (mental-health difficulties scores corrected for baseline) were regressed on 18 demographic, case and severity characteristics at baseline. We compared the performance of GLMM trees with that of traditional GLMMs and random forests. Results: GLMM trees yielded modest predictive accuracy, with cross-validated multiple R values of .18 and .25. Predictive accuracy did not differ significantly from that of traditional GLMMs and random forests, while GLMM trees required evaluation of a lower number of variables. Conclusion: GLMM trees provide a useful data-analytic tool for clinical prediction problems. The supplemental material provides a tutorial for replicating the GLMM tree analyses in R.

Machine-Learning vs. Expert-Opinion Driven Logistic Regression Modelling for Predicting 30-Day Unplanned Rehospitalisation in Preterm Babies: A Prospective, Population-Based Study (EPIPAGE 2)

Introduction: Preterm babies are a vulnerable population that experience significant short and long-term morbidity. Rehospitalisations constitute an important, potentially modifiable adverse event in this population. Improving the ability of clinicians to identify those patients at the greatest risk of rehospitalisation has the potential to improve outcomes and reduce costs. Machine-learning algorithms can provide potentially advantageous methods of prediction compared to conventional approaches like logistic regression. Objective: To compare two machine-learning methods (least absolute shrinkage and selection operator (LASSO) and random forest) to expert-opinion driven logistic regression modelling for predicting unplanned rehospitalisation within 30 days in a large French cohort of preterm babies. Design, Setting and Participants: This study used data derived exclusively from the population-based prospective cohort study of French preterm babies, EPIPAGE 2. Only those babies discharged home alive and whose parents completed the 1-year survey were eligible for inclusion in our study. All predictive models used a binary outcome, denoting a baby's status for an unplanned rehospitalisation within 30 days of discharge. Predictors included those quantifying clinical, treatment, maternal and socio-demographic factors. The predictive abilities of models constructed using LASSO and random forest algorithms were compared with a traditional logistic regression model. The logistic regression model comprised 10 predictors, selected by expert clinicians, while the LASSO and random forest included 75 predictors. Performance measures were derived using 10-fold cross-validation. Performance was quantified using area under the receiver operator characteristic curve, sensitivity, specificity, Tjur's coefficient of determination and calibration measures. Results: The rate of 30-day unplanned rehospitalisation in the eligible population used to construct the models was 9.1% (95% CI 8.2-10.1) (350/3,841). The random forest model demonstrated both an improved AUROC (0.65; 95% CI 0.59-0.7; p = 0.03) and specificity vs. logistic regression (AUROC 0.57; 95% CI 0.51-0.62, p = 0.04). The LASSO performed similarly (AUROC 0.59; 95% CI 0.53-0.65; p = 0.68) to logistic regression. Conclusions: Compared to an expert-specified logistic regression model, random forest offered improved prediction of 30-day unplanned rehospitalisation in preterm babies. However, all models offered relatively low levels of predictive ability, regardless of modelling method.

Personalised analytics for rare disease diagnostics

Whole genome and exome sequencing is a standard tool for the diagnosis of patients suffering from rare and other genetic disorders. The interpretation of the tens of thousands of variants returned from such tests remains a major challenge. Here we focus on the problem of prioritising variants with respect to the observed disease phenotype. We hypothesise that linking patterns of gene expression across multiple tissues to the phenotypes will aid in discovering disease causing variants. To test this, we construct classifiers that learn associations between tissue-specific gene expression and disease phenotypes. We find that using Genotype-Tissue Expression project (GTEx) expression data in conjunction with disease agnostic variant prioritisation methods (CADD or MetaSVM) results in consistent improvements in classification accuracy. Our method represents a previously overlooked avenue of utilising existing expression data for clinical diagnostics, and also opens the door to use of other functional genomic data sets in the same manner.

Exploration of Cervical Myelopathy Location From Somatosensory Evoked Potentials Using Random Forests Classification

Studies using time-frequency analysis have reported that somatosensory evoked potentials provide information regarding the location of spinal cord injury. However, a better understanding of the time-frequency components derived from somatosensory evoked potentials is essential for developing more reliable algorithms that can diagnosis level (location) of cervical injury. In the present study, we proposed a random forests machine learning approach, for separating somatosensory evoked potentials depending on spinal cord state. For data acquisition, we established rat models of compression spinal cord injury at the C4, C5, and C6 levels to induce cervical myelopathy. After making the compression injury, we collected somatosensory evoked potentials and extracted their time-frequency components. We then used the random forests classification system to analyze the evoked potential dataset that was obtained from the three groups of model rats. Evaluation of the classifier performance revealed an overall classification accuracy of 84.72%, confirming that the random forests method was able to separate the time-frequency components of somatosensory evoked potentials from rats under different conditions. Features of the time-frequency components contained information that could identify the location of the cervical spinal cord injury, demonstrating the potential benefits of using time-frequency components of somatosensory evoked potentials to diagnose the level of cervical injury in cervical myelopathy.

Multimodal pain assessment improves discrimination between noxious and non‐noxious stimuli in infants

Infants in neonatal intensive care units frequently experience clinically necessary painful procedures, which elicit a range of behavioral, physiological, and neurophysiological responses. However, the measurement of pain in this population is a challenge and no gold standard exists. The aim of this study was to investigate how noxious‐evoked changes in facial expression, reflex withdrawal, brain activity, heart rate, and oxygen saturation are related and to examine their accuracy in discriminating between noxious and non‐noxious stimuli. In 109 infants who received a clinically required heel lance and a control non‐noxious stimulus, we investigated whether combining responses across each modality, or including multiple measures from within each modality improves our ability to discriminate the noxious and non‐noxious stimuli. A random forest algorithm was used to build data‐driven models to discriminate between the noxious and non‐noxious stimuli in a training set which were then validated in a test set of independent infants. Measures within each modality were highly correlated, while different modalities showed less association. The model combining information across all modalities had good discriminative ability (accuracy of 0.81 in identifying noxious and non‐noxious stimuli), which was higher than the discriminative power of the models built from individual modalities. This demonstrates the importance of including multiple modalities in the assessment of infant pain.

HOME: a histogram based machine learning approach for effective identification of differentially methylated regions

Background

The development of whole genome bisulfite sequencing has made it possible to identify methylation differences at single base resolution throughout an entire genome. However, a persistent challenge in DNA methylome analysis is the accurate identification of differentially methylated regions (DMRs) between samples. Sensitive and specific identification of DMRs among different conditions requires accurate and efficient algorithms, and while various tools have been developed to tackle this problem, they frequently suffer from inaccurate DMR boundary identification and high false positive rate.

Results

We present a novel Histogram Of MEthylation (HOME) based method that takes into account the inherent difference in the distribution of methylation levels between DMRs and non-DMRs to discriminate between the two using a Support Vector Machine. We show that generated features used by HOME are dataset-independent such that a classifier trained on, for example, a mouse methylome training set of regions of differentially accessible chromatin, can be applied to any other organism’s dataset and identify accurate DMRs. We demonstrate that DMRs identified by HOME exhibit higher association with biologically relevant genes, processes, and regulatory events compared to the existing methods. Moreover, HOME provides additional functionalities lacking in most of the current DMR finders such as DMR identification in non-CG context and time series analysis. HOME is freely available at https://github.com/ListerLab/HOME.

Conclusion

HOME produces more accurate DMRs than the current state-of-the-art methods on both simulated and biological datasets. The broad applicability of HOME to identify accurate DMRs in genomic data from any organism will have a significant impact upon expanding our knowledge of how DNA methylation dynamics affect cell development and differentiation.

Electronic supplementary material

The online version of this article (10.1186/s12859-019-2845-y) contains supplementary material, which is available to authorized users.

Generating Tree-Level Harvest Predictions from Forest Inventories with Random Forests

Wood supply predictions from forest inventories involve two steps. First, it is predicted whether harvests occur on a plot in a given time period. Second, for plots on which harvests are predicted to occur, the harvested volume is predicted. This research addresses this second step. For forests with more than one species and/or forests with trees of varying dimensions, overall harvested volume predictions are not satisfactory and more detailed predictions are required. The study focuses on southwest Germany where diverse forest types are found. Predictions are conducted for plots on which harvests occurred in the 2002–2012 period. For each plot, harvest probabilities of sample trees are predicted and used to derive the harvested volume (m³ over bark in 10 years) per hectare. Random forests (RFs) have become popular prediction models as they define the interactions and relationships of variables in an automatized way. However, their suitability for predicting harvest probabilities for inventory sample trees is questionable and has not yet been examined. Generalized linear mixed models (GLMMs) are suitable in this context as they can account for the nested structure of tree-level data sets (trees nested in plots). It is unclear if RFs can cope with this data structure. This research aims to clarify this question by comparing two RFs—an RF based on conditional inference trees (CTree-RF), and an RF based on classification and regression trees (CART-RF)—with a GLMM. For this purpose, the models were fitted on training data and evaluated on an independent test set. Both RFs achieved better prediction results than the GLMM. Regarding plot-level harvested volumes per ha, they achieved higher variances explained (VEs) and significantly (p < 0.05) lower mean absolute residuals when compared to the GLMM. VEs were 0.38 (CTree-RF), 0.37 (CART-RF), and 0.31 (GLMM). Root means squared errors were 138.3, 139.9 and 145.5, respectively. The research demonstrates the suitability and advantages of RFs for predicting harvest decisions on the level of inventory sample trees. RFs can become important components within the generation of business-as-usual wood supply scenarios worldwide as they are able to learn and predict harvest decisions from NFIs in an automatized and self-adapting way. The applied approach is not restricted to specific forests or harvest regimes and delivers detailed species and dimension information for the harvested volumes.

Radiomics of Brain MRI: Utility in Prediction of Metastatic Tumor Type

Purpose To investigate the feasibility of tumor type prediction with MRI radiomic image features of different brain metastases in a multiclass machine learning approach for patients with unknown primary lesion at the time of diagnosis. Materials and methods This single-center retrospective analysis included radiomic features of 658 brain metastases from T1-weighted contrast material-enhanced, T1-weighted nonenhanced, and fluid-attenuated inversion recovery (FLAIR) images in 189 patients (101 women, 88 men; mean age, 61 years; age range, 32-85 years). Images were acquired over a 9-year period (from September 2007 through December 2016) with different MRI units, reflecting heterogeneous image data. Included metastases originated from breast cancer (n = 143), small cell lung cancer (n = 151), non-small cell lung cancer (n = 225), gastrointestinal cancer (n = 50), and melanoma (n = 89). A total of 1423 quantitative image features and basic clinical data were evaluated by using random forest machine learning algorithms. Validation was performed with model-external fivefold cross validation. Comparative analysis of 10 randomly drawn cross-validation sets verified the stability of the results. The classifier performance was compared with predictions from a respective conventional reading by two radiologists. Results Areas under the receiver operating characteristic curve of the five-class problem ranged between 0.64 (for non-small cell lung cancer) and 0.82 (for melanoma); all P values were less than .01. Prediction performance of the classifier was superior to the radiologists' readings. Highest differences were observed for melanoma, with a 17-percentage-point gain in sensitivity compared with the sensitivity of both readers; P values were less than .02. Conclusion Quantitative features of routine brain MR images used in a machine learning classifier provided high discriminatory accuracy in predicting the tumor type of brain metastases. © RSNA, 2018 Online supplemental material is available for this article.

Predicting in vivo effect levels for repeat-dose systemic toxicity using chemical, biological, kinetic and study covariates

In an effort to address a major challenge in chemical safety assessment, alternative approaches for characterizing systemic effect levels, a predictive model was developed. Systemic effect levels were curated from ToxRefDB, HESS-DB and COSMOS-DB from numerous study types totaling 4379 in vivo studies for 1247 chemicals. Observed systemic effects in mammalian models are a complex function of chemical dynamics, kinetics, and inter- and intra-individual variability. To address this complex problem, systemic effect levels were modeled at the study-level by leveraging study covariates (e.g., study type, strain, administration route) in addition to multiple descriptor sets, including chemical (ToxPrint, PaDEL, and Physchem), biological (ToxCast), and kinetic descriptors. Using random forest modeling with cross-validation and external validation procedures, study-level covariates alone accounted for approximately 15% of the variance reducing the root mean squared error (RMSE) from 0.96 log10 to 0.85 log10 mg/kg/day, providing a baseline performance metric (lower expectation of model performance). A consensus model developed using a combination of study-level covariates, chemical, biological, and kinetic descriptors explained a total of 43% of the variance with an RMSE of 0.69 log10 mg/kg/day. A benchmark model (upper expectation of model performance) was also developed with an RMSE of 0.5 log10 mg/kg/day by incorporating study-level covariates and the mean effect level per chemical. To achieve a representative chemical-level prediction, the minimum study-level predicted and observed effect level per chemical were compared reducing the RMSE from 1.0 to 0.73 log10 mg/kg/day, equivalent to 87% of predictions falling within an order-of-magnitude of the observed value. Although biological descriptors did not improve model performance, the final model was enriched for biological descriptors that indicated xenobiotic metabolism gene expression, oxidative stress, and cytotoxicity, demonstrating the importance of accounting for kinetics and non-specific bioactivity in predicting systemic effect levels. Herein, we generated an externally predictive model of systemic effect levels for use as a safety assessment tool and have generated forward predictions for over 30,000 chemicals.

The fecal microbiota as a biomarker for disease activity in Crohn's disease

Monitoring mucosal inflammation is crucial to prevent complications and disease progression in Crohn's disease (CD). Endoscopy is the current standard, but is invasive. Clinical activity scores and non-invasive biochemical markers do not correlate well with mucosal inflammation. Microbial perturbations have been associated with disease activity in CD. Therefore, we aimed to investigate its potential use to differentiate CD patients in remission from those with an exacerbation. From 71 CD patients repeated fecal samples were collected, resulting in 97 active disease and 97 remission samples based on a combination of biochemical and clinical parameters. The microbiota composition was assessed by pyrosequencing of the 16S rRNA V1-V3 region. Random Forest analysis was used to find the most discriminatory panel of operational taxonomic units (OTUs) between active and remission samples. An independent internal validation set was used to validate the model. A combination of 50 OTUs was able to correctly predict 73% of remission and 79% of active samples with an AUC of 0.82 (sensitivity: 0.79, specificity: 0.73). This study demonstrates that fecal microbial profiles can be used to differentiate between active and remission CD and underline the potential of the fecal microbiota as a non-invasive tool to monitor disease activity in CD.

The Random Forests statistical technique: An examination of its value for the study of reading

Studies investigating individual differences in reading ability often involve data sets containing a large number of collinear predictors and a small number of observations. In this paper, we discuss the method of Random Forests and demonstrate its suitability for addressing the statistical concerns raised by such datasets. The method is contrasted with other methods of estimating relative variable importance, especially Dominance Analysis and Multimodel Inference. All methods were applied to a dataset that gauged eye-movements during reading and offline comprehension in the context of multiple ability measures with high collinearity due to their shared verbal core. We demonstrate that the Random Forests method surpasses other methods in its ability to handle model overfitting, and accounts for a comparable or larger amount of variance in reading measures relative to other methods.

Gene-Gene Interaction Among WNT Genes for Oral Cleft in Trios

Genome-wide association studies (GWAS) for nonsyndromic cleft lip with or without cleft palate (CL/P) have identified multiple genes as important in the etiology of this common birth defect. We performed a candidate gene/pathway analysis explicitly considering gene-gene (G × G) interaction to further explore the etiology of CL/P. Animal models have shown the WNT signaling pathway plays an important role in mid-facial development, and various genes in this pathway have been associated with nonsyndromic CL/P in previous studies. We propose a combined approach to search for possible G × G interactions using machine learning and regression-based methods to test for interactions between genes in the WNT family, and between these genes and other genes identified by GWAS in case-parent trios. Using this combined approach of regression-based and machine learning methods in CL/P case-parent trios, we found robust evidence of G × G interaction between markers in WNT5B and MAFB (empiric P-values = 0.0076 among Asian trios and P-values = 0.018 among European trios). Additional evidence for epistatic interaction between markers in WNT5A, IRF6, and C1orf107 was seen among Asian trios, and markers in the 8q24 region and WNT5B among European trios.

Prediction of remission in obsessive compulsive disorder using a novel machine learning strategy

The study objective was to apply machine learning methodologies to identify predictors of remission in a longitudinal sample of 296 adults with a primary diagnosis of obsessive compulsive disorder (OCD). Random Forests is an ensemble machine learning algorithm that has been successfully applied to large-scale data analysis across vast biomedical disciplines, though rarely in psychiatric research or for application to longitudinal data. When provided with 795 raw and composite scores primarily from baseline measures, Random Forest regression prediction explained 50.8% (5000-run average, 95% bootstrap confidence interval [CI]: 50.3-51.3%) of the variance in proportion of time spent remitted. Machine performance improved when only the most predictive 24 items were used in a reduced analysis. Consistently high-ranked predictors of longitudinal remission included Yale-Brown Obsessive Compulsive Scale (Y-BOCS) items, NEO items and subscale scores, Y-BOCS symptom checklist cleaning/washing compulsion score, and several self-report items from social adjustment scales. Random Forest classification was able to distinguish participants according to binary remission outcomes with an error rate of 24.6% (95% bootstrap CI: 22.9-26.2%). Our results suggest that clinically-useful prediction of remission may not require an extensive battery of measures. Rather, a small set of assessment items may efficiently distinguish high- and lower-risk patients and inform clinical decision-making.

Profiling of volatile organic compounds in exhaled breath as a strategy to find early predictive signatures of asthma in children

Wheezing is one of the most common respiratory symptoms in preschool children under six years old. Currently, no tests are available that predict at early stage who will develop asthma and who will be a transient wheezer. Diagnostic tests of asthma are reliable in adults but the same tests are difficult to use in children, because they are invasive and require active cooperation of the patient. A non-invasive alternative is needed for children. Volatile Organic Compounds (VOCs) excreted in breath could yield such non-invasive and patient-friendly diagnostic. The aim of this study was to identify VOCs in the breath of preschool children (inclusion at age 2-4 years) that indicate preclinical asthma. For that purpose we analyzed the total array of exhaled VOCs with Gas Chromatography time of flight Mass Spectrometry of 252 children between 2 and 6 years of age. Breath samples were collected at multiple time points of each child. Each breath-o-gram contained between 300 and 500 VOCs; in total 3256 different compounds were identified across all samples. Using two multivariate methods, Random Forests and dissimilarity Partial Least Squares Discriminant Analysis, we were able to select a set of 17 VOCs which discriminated preschool asthmatic children from transient wheezing children. The correct prediction rate was equal to 80% in an independent test set. These VOCs are related to oxidative stress caused by inflammation in the lungs and consequently lipid peroxidation. In conclusion, we showed that VOCs in the exhaled breath predict the subsequent development of asthma which might guide early treatment.

Bacterial profile of dentine caries and the impact of pH on bacterial population diversity

Dental caries is caused by the release of organic acids from fermentative bacteria, which results in the dissolution of hydroxyapatite matrices of enamel and dentine. While low environmental pH is proposed to cause a shift in the consortium of oral bacteria, favouring the development of caries, the impact of this variable has been overlooked in microbial population studies. This study aimed to detail the zonal composition of the microbiota associated with carious dentine lesions with reference to pH. We used 454 sequencing of the 16S rRNA gene (V3–V4 region) to compare microbial communities in layers ranging in pH from 4.5–7.8 from 25 teeth with advanced dentine caries. Pyrosequencing of the amplicons yielded 449,762 sequences. Nine phyla, 97 genera and 409 species were identified from the quality-filtered, de-noised and chimera-free sequences. Among the microbiota associated with dentinal caries, the most abundant taxa included Lactobacillus sp., Prevotella sp., Atopobium sp., Olsenella sp. and Actinomyces sp. We found a disparity between microbial communities localised at acidic versus neutral pH strata. Acidic conditions were associated with low diversity microbial populations, with Lactobacillus species including L. fermentum, L. rhamnosus and L. crispatus, being prominent. In comparison, the distinctive species of a more diverse flora associated with neutral pH regions of carious lesions included Alloprevotella tanerrae, Leptothrix sp., Sphingomonas sp. and Streptococcus anginosus. While certain bacteria were affected by the pH gradient, we also found that ∼60% of the taxa associated with caries were present across the investigated pH range, representing a substantial core. We demonstrated that some bacterial species implicated in caries progression show selective clustering with respect to pH gradient, providing a basis for specific therapeutic strategies.

Travelling the world of gene-gene interactions

Over the last few years, main effect genetic association analysis has proven to be a successful tool to unravel genetic risk components to a variety of complex diseases. In the quest for disease susceptibility factors and the search for the 'missing heritability', supplementary and complementary efforts have been undertaken. These include the inclusion of several genetic inheritance assumptions in model development, the consideration of different sources of information, and the acknowledgement of disease underlying pathways of networks. The search for epistasis or gene-gene interaction effects on traits of interest is marked by an exponential growth, not only in terms of methodological development, but also in terms of practical applications, translation of statistical epistasis to biological epistasis and integration of omics information sources. The current popularity of the field, as well as its attraction to interdisciplinary teams, each making valuable contributions with sometimes rather unique viewpoints, renders it impossible to give an exhaustive review of to-date available approaches for epistasis screening. The purpose of this work is to give a perspective view on a selection of currently active analysis strategies and concerns in the context of epistasis detection, and to provide an eye to the future of gene-gene interaction analysis.