Empirical Bayes Model Research Articles

Altered effective connectivity among face-processing systems in major depressive disorder.

Neuroimaging studies have revealed abnormal functional interaction during the processing of emotional faces in patients with major depressive disorder (MDD), thereby enhancing our comprehension of the pathophysiology of MDD. However, it is unclear whether there is abnormal directional interaction among face-processing systems in patients with MDD. A group of patients with MDD and a healthy control group underwent a face-matching task during functional magnetic resonance imaging. Dynamic causal modelling (DCM) analysis was used to investigate effective connectivity between 7 regions in the face-processing systems. We used a Parametric Empirical Bayes model to compare effective connectivity between patients with MDD and controls. We included 48 patients and 44 healthy controls in our analyses. Both groups showed higher accuracy and faster reaction time in the shape-matching condition than in the face-matching condition. However, no significant behavioural or brain activation differences were found between the groups. Using DCM, we found that, compared with controls, patients with MDD showed decreased self-connection in the right dorsolateral prefrontal cortex (DLPFC), amygdala, and fusiform face area (FFA) across task conditions; increased intrinsic connectivity from the right amygdala to the bilateral DLPFC, right FFA, and left amygdala, suggesting an increased intrinsic connectivity centred in the amygdala in the right side of the face-processing systems; both increased and decreased positive intrinsic connectivity in the left side of the face-processing systems; and comparable task modulation effect on connectivity. Our study did not include longitudinal neuroimaging data, and there was limited region of interest selection in the DCM analysis. Our findings provide evidence for a complex pattern of alterations in the face-processing systems in patients with MDD, potentially involving the right amygdala to a greater extent. The results confirm some previous findings and highlight the crucial role of the regions on both sides of face-processing systems in the pathophysiology of MDD.

New Empirical Bayes Models to Jointly Analyze Multiple RNA-Sequencing Data in a Hypophosphatasia Disease Study.

Hypophosphatasia is a rare inherited metabolic disorder caused by the deficiency of tissue-nonspecific alkaline phosphatase. More severe and early onset cases present symptoms of muscle weakness, diminished motor coordination, and epileptic seizures. These neurological manifestations are poorly characterized. Thus, it is urgent to discover novel differentially expressed genes for investigating the genetic mechanisms underlying the neurological manifestations of hypophosphatasia. RNA-sequencing data offer a high-resolution and highly accurate transcript profile. In this study, we apply an empirical Bayes model to RNA-sequencing data acquired from the spinal cord and neocortex tissues of a mouse model, individually, to more accurately estimate the genetic effects without bias. More importantly, we further develop two integration methods, weighted gene approach and weighted Z method, to incorporate two RNA-sequencing data into a model for enhancing the effects of genetic markers in the diagnostics of hypophosphatasia disease. The simulation and real data analysis have demonstrated the effectiveness of our proposed integration methods, which can maximize genetic signals identified from the spinal cord and neocortex tissues, minimize the prediction error, and largely improve the prediction accuracy in risk prediction.

Effective connectivity of the locus coeruleus in patients with late-life Major Depressive Disorder or mild cognitive impairment

IntroductionWe compared effective connectivity from the locus coeruleus (LC) during the resting-state in patients with late-life Major Depressive Disorder (MDD), individuals with amnestic Mild Cognitive Impairment (aMCI), and Healthy Controls (HCs). Participants23 patients with late-life MDD, 22 patients with aMCI, and 28 HCs. Material and methodsParticipants were assessed in two time-points, 2 years apart. They underwent a resting-state functional magnetic resonance imaging and a high-resolution anatomical acquisition, as well as clinical assessments. Functional imaging data were analyzed with dynamic causal modeling, and parametric empirical Bayes model was used to map effective connectivity between 7 distinct nodes: 4 from the locus coeruleus and 3 regions displaying gray matter decreases during the two-year follow-up period. ResultsLongitudinal analysis of structural data identified three clusters of larger over-time gray matter volume reduction in patients (MDD+aMCI vs. HCs): the right precuneus, and the visual association and parahippocampal cortices. aMCI patients showed decreased effective connectivity from the left rostral to caudal portions of the LC, while connectivity from the left rostral LC to the parahippocampal cortex increased. In MDD, there was a decline in effective connectivity across LC caudal seeds, and increased connectivity from the left rostral to the left caudal LC seed over time. Connectivity alterations with cortical regions involved cross-hemisphere increases and same-hemisphere decreases. ConclusionsOur discoveries provide insight into the dynamic changes in effective connectivity in individuals with late-life MDD and aMCI, also shedding light on the mechanisms potentially contributing to the onset of neurodegenerative disorders.

Nonparametric Estimation of Multivariate Copula Using Empirical Bayes Methods

In the fields of finance, insurance, system reliability, etc., it is often of interest to measure the dependence among variables by modeling a multivariate distribution using a copula. The copula models with parametric assumptions are easy to estimate but can be highly biased when such assumptions are false, while the empirical copulas are nonsmooth and often not genuine copulas, making the inference about dependence challenging in practice. As a compromise, the empirical Bernstein copula provides a smooth estimator, but the estimation of tuning parameters remains elusive. The proposed empirical checkerboard copula within a hierarchical empirical Bayes model alleviates the aforementioned issues and provides a smooth estimator based on multivariate Bernstein polynomials that itself is shown to be a genuine copula. Additionally, the proposed copula estimator is shown to provide a more accurate estimate of several multivariate dependence measures. Both theoretical asymptotic properties and finite-sample performances of the proposed estimator based on simulated data are presented and compared with some nonparametric estimators. An application to portfolio risk management is included based on stock prices data.

Clinical Correlates of Longitudinal Hypothalamic-Pontine Functional Changes in Migraine Patients (S41.008)

<h3>Objective:</h3> To explore longitudinal hypothalamic-pontine resting state (RS) effective connectivity (EC) changes in migraine patients and investigate their association with disease progression over the years. <h3>Background:</h3> The hypothalamus and pons play a pivotal role in migraine pathophysiology. Whether longitudinal hypothalamic-pontine RS EC changes could influence migraine disease activity has never been investigated. <h3>Design/Methods:</h3> RS functional magnetic resonance imaging were acquired from 92 headache-free episodic migraine patients and 73 controls. Twenty-three migraine patients and 23 controls were reexamined after 4 years. RS EC of bilateral pons and hypothalamus was performed using SPM12 and dynamic causal modelling. Longitudinal RS EC differences between groups were investigated using parametric empirical bayes models. <h3>Results:</h3> During the follow-up, 35% of patients reported an increased migraine attack frequency. Over the follow-up, migraine patients developed a higher inhibitory EC within the left pons, from the right pons to the ipsilateral hypothalamus, from the left hypothalamus to the ipsilateral and contralateral pons, and from the left pons to the ipsilateral and contralateral hypothalamus. Migraine patients experienced also a higher excitatory EC from the right pons to the left pons. At follow-up, greater headache impact correlated to higher inhibitory EC from the left pons to the ipsilateral and contralateral hypothalamus, while higher migraine attack frequency was associated with higher inhibitory left hypothalamic-left pontine EC. During the follow-up, the increased left pontine-right hypothalamic inhibitory EC was significantly associated to an increased migraine attack frequency. Lower inhibitory RS EC from the left pons to the left and right hypothalamus at baseline predicted clinical worsening over the years. <h3>Conclusions:</h3> Interictal migraine patients experience a prominent inhibitory influence of the pons over the hypothalamus and <i>vice versa</i> that could affect migraine progression over the years. An altered hypothalamic-pontine inhibitory activity may be a prognostic marker for migraine worsening after 4 years. <b>Disclosure:</b> Dr. Messina has received personal compensation in the range of $0-$499 for serving on a Scientific Advisory or Data Safety Monitoring board for Lundbeck . Dr. Messina has received personal compensation in the range of $0-$499 for serving on a Speakers Bureau for Eli Lilly. Dr. Messina has received personal compensation in the range of $0-$499 for serving on a Speakers Bureau for Bromatech. Paola Valsasina has received personal compensation in the range of $500-$4,999 for serving on a Speakers Bureau for ACCMED. Dr. Zanandrea has nothing to disclose. Dr. Cetta has nothing to disclose. Bruno Colombo has nothing to disclose. Maria Assunta Rocca has received personal compensation in the range of $500-$4,999 for serving on a Speakers Bureau for Bayer, Biogen, Bristol Myers Squibb, Celgene, Genzyme, Merck Serono, Novartis, Roche, and Teva. The institution of Maria Assunta Rocca has received research support from Italian Ministry of Health, MS Society of Canada and Fondazione Italiana Sclerosi Multipla. Dr. Filippi has received personal compensation in the range of $5,000-$9,999 for serving as a Consultant for Bayer, Biogen Idec, Merck-Serono, Novartis, Roche, Sanofi Genzyme, Takeda, and Teva Pharmaceutical Industries. Dr. Filippi has received personal compensation in the range of $5,000-$9,999 for serving on a Speakers Bureau for Bayer, Biogen Idec, Merck-Serono, Novartis, Roche, Sanofi Genzyme, Takeda, and Teva Pharmaceutical Industries. Dr. Filippi has received personal compensation in the range of $5,000-$9,999 for serving as an Editor, Associate Editor, or Editorial Advisory Board Member for Springer Nature. The institution of Dr. Filippi has received research support from Biogen Idec, Merck-Serono, Novartis, Roche, Teva Pharmaceutical Industries, Italian Ministry of Health, Fondazione Italiana Sclerosi Multipla, and ARiSLA (Fondazione Italiana di Ricerca per la SLA).

Discovery of associative patterns between workplace sound level and physiological wellbeing using wearable devices and empirical Bayes modeling

We conducted a field study using multiple wearable devices on 231 federal office workers to assess the impact of the indoor environment on individual wellbeing. Past research has established that the workplace environment is closely tied to an individual’s wellbeing. Since sound is the most-reported environmental factor causing stress and discomfort, we focus on quantifying its association with physiological wellbeing. Physiological wellbeing is represented as a latent variable in an empirical Bayes model with heart rate variability measures—SDNN and normalized-HF as the observed outcomes and with exogenous factors including sound level as inputs. We find that an individual’s physiological wellbeing is optimal when sound level in the workplace is at 50 dBA. At lower (<50dBA) and higher (>50dBA) amplitude ranges, a 10 dBA increase in sound level is related to a 5.4% increase and 1.9% decrease in physiological wellbeing respectively. Age, body-mass-index, high blood pressure, anxiety, and computer use intensive work are person-level factors contributing to heterogeneity in the sound-wellbeing association.

Estimating the likelihood of roadway pluvial flood based on crowdsourced traffic data and depression-based DEM analysis

Abstract. Water ponding and pluvial flash flooding (PFF) on roadways can pose a significant risk to drivers. Furthermore, climate change, growing urbanization, increasing imperviousness, and aging stormwater infrastructure have increased the frequency of these events. Using physics-based models to predict pluvial flooding at the road segment scale requires notable terrain simplifications and detailed information that is often not available at fine scales (e.g., blockage of stormwater inlets). This brings uncertainty into the results, especially in highly urbanized areas where micro-topographic features typically govern the actual flow dynamics. This study evaluates the potential for flood observations collected from Waze – a community-based navigation app – to estimate the likelihood of PFF at the road segment scale. We investigated the correlation of the Waze flood reports with well-known flood observations and maps, including the National Flood Hazard Layer (NFHL), high watermarks, and low water crossings data inventories. In addition, highly localized surface depressions and their catchments are derived from a 1 m resolution bare-earth digital elevation model (BE-DEM) to investigate the spatial association of Waze flood reports. This analysis showed that the highest correlation of Waze flood reports exists with local surface depressions rather than river flooding, indicating that they are potentially useful indicators of PFF. Accordingly, two data-driven models, empirical Bayes (EB) and random forest (RF) regression, were developed to predict the frequency of flooding, a proxy for flood susceptibility, for three classes of historical storm events (light, moderate, and severe) in every road segment with surface depressions. Applying the models to Waze data from 150 storms in the city of Dallas showed that depression catchment drainage area and imperviousness are the most important predictive features. The EB model performed with reasonable precision in estimating the number of PFF events out of 92 light, 41 moderate, and 17 severe storms with 0.84, 0.85, and 1.09 mean absolute errors, respectively. This study shows that Waze data provide useful information for highly localized PFF prediction. The superior performance of EB compared to the RF model shows that the historical observations included in the EB approach are important for more accurate PFF prediction.

Development of an imputation model to recalibrate birth weights measured in the early neonatal period to time at delivery and assessment of its impact on size-for-gestational age and low birthweight prevalence estimates: a secondary analysis of a pregnancy cohort in rural Nepal

ObjectivesIn low-income countries, birth weights for home deliveries are often measured at the nadir when babies may lose up of 10% of their birth weight, biasing estimates of small-for-gestational age...

Resting state effective connectivity abnormalities of the Papez circuit and cognitive performance in multiple sclerosis.

The Papez circuit is central to memory and emotional processes. However, little is known about its involvement in multiple sclerosis (MS). We aimed to investigate abnormalities of resting state (RS) effective connectivity (EC) between regions of the Papez circuit in MS and their relationship with cognitive performances. Sixty-two MS patients and 64 healthy controls (HC) underwent neuropsychological assessment, 3D T1-weighted, and RS functional MRI. RS EC analysis was performed using SPM12 and dynamic causal modeling. RS EC abnormalities were investigated using parametric empirical Bayes models and were correlated with cognitive scores. Compared to HC, MS patients showed (posterior probability > 0.95) higher EC between the right entorhinal cortex and right subiculum, and lower EC from the anterior cingulate cortex (ACC) to the posterior cingulate cortex (PCC), from left to right subiculum, from left anterior thalamus to ACC, and within ACC and PCC. Lower RS EC from the ACC to the PCC correlated with worse global cognitive scores (rho = 0.19; p = 0.03), worse visuospatial memory (rho = 0.19; p = 0.03) and worse semantic fluency (rho = 0.21; p = 0.02). Lower RS EC from the left to the right subiculum correlated with worse verbal memory (rho = 0.20; p = 0.02), lower RS EC within the ACC correlated with worse attention (rho = -0.19; p = 0.04) and more severe brain atrophy (rho = -0.26; p = 0.003). Higher EC from the right entorhinal cortex to right subiculum correlated with worse semantic fluency (rho = 0.21; p = 0.02). In conclusion, MS patients showed altered RS EC within the Papez circuit. Abnormal RS EC involving cingulate cortices and hippocampal formation contributed to explain cognitive deficits.

Numerical characterization of support recovery in sparse regression with correlated design

Sparse regression is employed in diverse scientific settings as a feature selection method. A pervasive aspect of scientific data is the presence of correlations between predictive features. These correlations hamper both feature selection and estimation and jeopardize conclusions drawn from estimated models. On the other hand, theoretical results on sparsity-inducing regularized regression have largely addressed conditions for selection consistency via asymptotics, and disregard the problem of model selection, whereby regularization parameters are chosen. In this numerical study, we address these issues through exhaustive characterization of the performance of several regression estimators, coupled with a range of model selection strategies. These estimators and selection criteria were examined across correlated regression problems with varying degrees of signal to noise, distributions of non-zero model coefficients, and model sparsity. Our results reveal a fundamental tradeoff between false positive and false negative control in all regression estimators and model selection criteria examined. Additionally, we numerically explore a transition point modulated by the signal-to-noise ratio and spectral properties of the design covariance matrix at which the selection accuracy of all considered algorithms degrades. Overall, we find that SCAD coupled with BIC or empirical Bayes model selection performs the best feature selection across the regression problems considered.

Model-based identification of conditionally-essential genes from transposon-insertion sequencing data.

The understanding of bacterial gene function has been greatly enhanced by recent advancements in the deep sequencing of microbial genomes. Transposon insertion sequencing methods combines next-generation sequencing techniques with transposon mutagenesis for the exploration of the essentiality of genes under different environmental conditions. We propose a model-based method that uses regularized negative binomial regression to estimate the change in transposon insertions attributable to gene-environment changes in this genetic interaction study without transformations or uniform normalization. An empirical Bayes model for estimating the local false discovery rate combines unique and total count information to test for genes that show a statistically significant change in transposon counts. When applied to RB-TnSeq (randomized barcode transposon sequencing) and Tn-seq (transposon sequencing) libraries made in strains of Caulobacter crescentus using both total and unique count data the model was able to identify a set of conditionally beneficial or conditionally detrimental genes for each target condition that shed light on their functions and roles during various stress conditions.

Empirical Bayes Model Averaging with Influential Observations: Tuning Zellner’s [formula omitted] Prior for Predictive Robustness

The behavior of Bayesian model averaging (BMA) for the normal linear regression model in the presence of influential observations that contribute to model misfit is investigated. Remedies to attenuate the potential negative impacts of such observations on inference and prediction are proposed. The methodology is motivated by the view that well-behaved residuals and good predictive performance often go hand-in-hand. Focus is placed on regression models that use variants on Zellner's g prior. Studying the impact of various forms of model misfit on BMA predictions in simple situations points to prescriptive guidelines for “tuning” Zellner's g prior to obtain optimal predictions. The tuning of the prior distribution is obtained by considering theoretical properties that should be enjoyed by the optimal fits of the various models in the BMA ensemble. The methodology can be thought of as an “empirical Bayes” approach to modeling, as the data help to inform the specification of the prior in an attempt to attenuate the negative impact of influential cases.

Causal Interactions Between the Default Mode Network and Central Executive Network in Patients with Major Depression

Two different but interacting neural systems exist in the human brain: the task positive networks and task negative networks. One of the most important task positive networks is the central executive network (CEN), while the task negative network generally refers to the default mode network (DMN), which usually demonstrates task-induced deactivation. Although previous studies have clearly shown the association of both the CEN and DMN with major depressive disorder (MDD), how the causal interactions between these two networks change in depressed patients remains unclear. In the current study, 99 subjects (43 patients with MDD and 56 healthy controls) were recruited with their resting-state fMRI data collected. After data preprocessing, spectral dynamic causal modeling (spDCM) was used to investigate the causal interactions within and between the DMN and CEN. Group commonalities and differences in causal interaction patterns within and between the CEN and DMN in patients and controls were assessed by a parametric empirical Bayes (PEB) model. Both subject groups demonstrated significant effective connectivity between regions of the CEN and DMN. In particular, we detected inhibitory influences from the CEN to the DMN with node-level PEB analyses, which may help to explain the anticorrelations between these two networks consistently reported in previous studies. Compared with healthy controls, patients with MDD showed increased effective connectivity within the CEN and decreased connectivity from regions of the CEN to DMN, suggesting impaired control of the DMN by the CEN in these patients. These findings might provide new insights into the neural substrates of MDD.

Inference on Reported Vehicular Fatal Accidents in Nigeria Using a Bayesian Model

The study introduced a special case of the Poisson-Generalized Gamma empirical Bayes model to survey states in Nigeria with a higher risk of fatal accidents. Monte Carlo error and stationary dynamic trace plots were used to validate model convergence and accuracy of the posterior estimates. The main results included the disease mappings that revealed Ebonyi had the highest risk of road vehicular fatal accidents in Nigeria with a relative risk estimate of 1.4120 while Abuja had the lowest risk with a relative risk estimate 0.5711. In terms of geopolitical region, the risk of road vehicular fatal accident is highest in South-South region with a relative risk estimate of 1.1850 while North-Central had the lowest risk with a relative risk estimate of 0.7846. The study is to aid planned government programs to ameliorate vehicular road carnage in Nigeria.

Bayesian Model for HOPE Mass Spectrometers on Van Allen Probes

AbstractSpace instruments rely heavily on modeling to predict and understand the instrument response, enabling a determination of the capabilities and resolution. The Bayesian approach provides a framework to incorporate prior knowledge and propagate uncertainty to predict the instrument response. We present an empirical Bayes model for the end‐to‐end performance of the Helium, Oxygen, Proton, and Electron mass spectrometers aboard the Van Allen Probes mission. In this model, we use a combination of external modeling, laboratory calibration, and expert opinion to construct the time‐of‐flight spectra and demonstrate good agreement with on‐orbit data. The empirical Bayes model is applied to explore doubly charged ions and carbon, nitrogen, oxygen discrimination during the Van Allen Probes mission.

Identifying outlier patterns of inconsistent ambulance billing in Medicare.

To illustrate a method that accounts for sampling variation in identifying suppliers and counties with outlying rates of a particular pattern of inconsistent billing for ambulance services to Medicare. US Medicare claims for a 20% simple random sample of 2010-2014 fee-for-service beneficiaries. We identified instances in which ambulance suppliers billed Medicare for transporting a patient to a hospital, but no corresponding hospital visit appeared in billing claims. We estimated the distributions of outlier supplier and county rates of such "ghost rides" by fitting a nonparametric empirical Bayes model with flexible distributional assumptions to account for sampling variation. We included Basic and advanced life support ground emergency ambulance claims with a hospital destination. "Ghost ride" rates varied considerably across both ambulance suppliers and counties. We estimated 6.1% of suppliers and 5.0% of counties had rates that exceeded 3.6%, which was twice the national average of "ghost rides" (1.8% of all ambulance transports). Health care fraud and abuse are frequently asserted but can be difficult to detect. Our data-driven approach may be a useful starting point for further investigation.

Review on diabetes disease management programs

To comprehensively review and integrate available data on the impact of diabetes treatment programs on care procedures and outcomes for patients with diabetes mellitus. Review of literature and morph were used in this study. From 1987 to 2001, computerized databases was conducted for English-language controlled trials evaluating the efficacy of diabetic disease management regimens. Using a standardized abstraction form, two independent reviewers collected research data. An empirical Bayes model was used to generate pooling estimates of program impacts on glycosylated hemoglobin. The pooled assessment of program effects on glycosylated hemoglobin was a 0.5-percentage-point decrease (95 percent confidence range, 0.3 to 0.6 percentage points), which is a small but substantial improvement. Research strongly supports the program's advantages in improving retinal and foot lesions screening. Diabetic programs may help to enhance glucose control and boost retinal and foot problems screening. For individuals with diabetes mellitus, greater effort will be needed to develop more successful disease management strategies.

Integrating Biological Knowledge Into Case-Control Analysis Through Iterated Conditional Modes/Medians Algorithm.

Logistic regression is an effective tool in case-control analysis. With the advanced high throughput technology, a quest to seek a fast and efficient method in fitting high-dimensional logistic regression has gained much interest. An empirical Bayes model for logistic regression is considered in this article. A spike-and-slab prior is used for variable selection purpose, which plays a vital role in building an effective predictive model while making model interpretable. To increase the power of variable selection, we incorporate biological knowledge through the Ising prior. The development of the iterated conditional modes/medians (ICM/M) algorithm is proposed to fit the logistic model that has computational advantage over Markov Chain Monte Carlo (MCMC) algorithms. The implementation of the ICM/M algorithm for both linear and logistic models can be found in R package icmm that is freely available on Comprehensive R Archive Network (CRAN). Simulation studies were carried out to assess the performances of our method, with lasso and adaptive lasso as benchmark. Overall, the simulation studies show that the ICM/M outperform the others in terms of number of false positives and have competitive predictive ability. An application to a real data set from Parkinson's disease study was also carried out for illustration. To identify important variables, our approach provides flexibility to select variables based on local posterior probabilities while controlling false discovery rate at a desired level rather than relying only on regression coefficients.

P2‐386: EFFECTIVE CONNECTIVITY WITHIN THE LEFT AND RIGHT EXECUTIVE CONTROL NETWORKS IN MCI AND AD

The default mode network, implicated in various types of memory, is often investigated in Alzheimer's disease (AD). AD patients often exhibit both memory impairments and executive dysfunction. The executive control network (ECN), thought to be linked with the default mode network and implicated in executive function, differs in functional connectivity, often measured using Pearson correlation, between healthy controls and individuals with AD (Agosta et al., 2012). Differences in effective connectivity, a measure of causal neuronal interaction, as measured using Granger causality were noted between healthy controls, individuals with mild cognitive impairment (MCI), and individuals with AD (Cai et al., 2017). ECN effective connectivity (EC) was explored here using dynamic causal modeling (DCM) to obtain EC at the unobserved neuronal level rather than the observed BOLD signal by accounting for local differences in hemodynamic response. Whole-brain simultaneous multi-slice resting state fMRI data were acquired on 3T GE 750 scanners at the University of Wisconsin-Madison and Medical College of Wisconsin (8 bands, 72 slices, TR 802 ms, voxel size 2 mm isotropic). Participants were enrolled in the NIH-sponsored Alzheimer's Disease Connectome Project (Table 1). Slice time correction was applied prior to the Human Connectome Project minimal preprocessing pipeline (Glasser et al., 2013). Separate fully connected spectral DCMs (Friston, Kahan, Biswal, and Razi, 2014) were constructed for the Stanford FIND lab's left and right ECN (Shirer, Ryali, Rykhlevskaia, Menon, and Greicius, 2012) following nuisance regression. Group differences were investigated via parametric empirical Bayes (PEB) models (Friston et al., 2016). Automated search removed non-evident parameters and a threshold of posterior probability of 95% was used. Three/nine parameters in the left ECN differed between controls and individuals with MCI/AD (Table 2). Eight/seven parameters in the right ECN differed between controls and individuals with MCI/AD (Table 3). All parameters in the left ECN which differed in MCI compared to controls also differed in AD while 6 difference parameters were in common in the right ECN.

Machine learning techniques combined with dose profiles indicate radiation response biomarkers

Abstract The focus of this research is to combine statistical and machine learning tools in application to a high-throughput biological data set on ionizing radiation response. The analyzed data consist of two gene expression sets obtained in studies of radiosensitive and radioresistant breast cancer patients undergoing radiotherapy. The data sets were similar in principle; however, the treatment dose differed. It is shown that introducing mathematical adjustments in data preprocessing, differentiation and trend testing, and classification, coupled with current biological knowledge, allows efficient data analysis and obtaining accurate results. The tools used to customize the analysis workflow were batch effect filtration with empirical Bayes models, identifying gene trends through the Jonckheere–Terpstra test and linear interpolation adjustment according to specific gene profiles for multiple random validation. The application of non-standard techniques enabled successful sample classification at the rate of 93.5% and the identification of potential biomarkers of radiation response in breast cancer, which were confirmed with an independent Monte Carlo feature selection approach and by literature references. This study shows that using customized analysis workflows is a necessary step towards novel discoveries in complex fields such as personalized individual therapy.