An integrated Bayesian framework for multi-omics prediction and classification.

With the growing commonality of multi-omics datasets, there is now increasing evidence that integrated omics profiles lead to more efficient discovery of clinically actionable biomarkers that enable better disease outcome prediction and patient stratification. Several methods exist to perform host phenotype prediction from cross-sectional, single-omics data modalities but decentralized frameworks that jointly analyze multiple time-dependent omics data to highlight the integrative and dynamic impact of repeatedly measured biomarkers are currently limited. In this article, we propose a novel Bayesian ensemble method to consolidate prediction by combining information across several longitudinal and cross-sectional omics data layers. Unlike existing frequentist paradigms, our approach enables uncertainty quantification in prediction as well as interval estimation for a variety of quantities of interest based on posterior summaries. We apply our method to four published multi-omics datasets and demonstrate that it recapitulates known biology in addition to providing novel insights while also outperforming existing methods in estimation, prediction, and uncertainty quantification. Our open-source software is publicly available at https://github.com/himelmallick/IntegratedLearner.

Use of the single-item Patient Global Impression-Severity scale as a self-reported assessment of insomnia severity.

We evaluated a single-item Patient Global Impression-Severity (PGI-S) scale for assessing insomnia severity during the clinical development programme for suvorexant. The analyses used data from two randomised, double-blind, placebo-controlled, 3-month, Phase III clinical trials of suvorexant in patients with Diagnostic and Statistical Manual of Mental Disorders IV criteria insomnia. Patients assessed insomnia severity during the previous week using the PGI-S, a one-item questionnaire containing six response options ranging from 0 (none) to 5 (very severe), at baseline and at Week 2, and Months 1, 2, and 3 after randomisation. The seven-item Insomnia Severity Index (ISI) and other subjective and objective assessments were also completed by patients. PGI-S responses were compared primarily with the ISI using descriptive statistics and correlations. The PGI-S demonstrated favourable measurement characteristics (validity, reliability, responsiveness and sensitivity). PGI-S scores decreased from baseline to Month 3 in a similar pattern to the ISI total score, and the Spearman correlation coefficient between PGI-S and the ISI was .73. An improvement of ≥2 points on the PGI-S defined a treatment responder, based on comparison to the ISI definition of a responder (improvement of ≥6 points). Our present findings suggest that the PGI-S is a simple but valid, reliable, responsive, sensitive, and meaningful patient-reported assessment of insomnia severity. The PGI-S may be particularly useful as a companion outcome to sleep monitoring using wearable sleep devices or smartphones in at-home settings.

Pilot evaluation of a consumer wearable device to assess sleep in a clinical polysomnography trial of suvorexant for treating insomnia in patients with Alzheimer's disease.

The orexin receptor antagonist suvorexant was previously reported to significantly improve total sleep time (TST), by 28 min per night versus placebo after 4 weeks, in a sleep laboratory polysomnography (PSG) study of patients with Alzheimer's disease and insomnia. The study included an exploratory evaluation of a consumer-grade wearable "watch" device for assessing sleep that we report on here. Participants who met diagnostic criteria for both probable Alzheimer's disease dementia and insomnia were randomized to suvorexant 10-20 mg (N = 142) or placebo (N = 143) in a double-blind, 4-week trial. Patients were provided with a consumer-grade wearable watch device (Garmin vívosmart® HR) to be worn continuously. Overnight sleep laboratory PSG was performed on three nights: screening, baseline and Night 29 (last dose). Watch treatment effects were assessed by change-from-baseline in watch TST at Week 4 (average TST per night). We also analysed Night 29 data only, with watch data restricted to the PSG recording time. In the 193 participants included in the Week 4 watch analysis (suvorexant = 97, placebo = 96), the suvorexant-placebo difference in watch TST was 4 min (p = .622). In patients with usable data for both assessments at the baseline and Night 29 PSG (suvorexant = 57, placebo = 50), the watch overestimated TST compared to PSG (e.g., placebo baseline = 412 min for watch and 265 min for PSG) and underestimated change-from-baseline treatment effects: the suvorexant-placebo difference was 20 min for watch TST (p = .405) and 35 min for PSG TST (p = .057). These findings show that the watch was less sensitive than PSG for evaluating treatment effects on TST.

The reciprocal Bayesian LASSO.

A reciprocal LASSO (rLASSO) regularization employs a decreasing penalty function as opposed to conventional penalization approaches that use increasing penalties on the coefficients, leading to stronger parsimony and superior model selection relative to traditional shrinkage methods. Here we consider a fully Bayesian formulation of the rLASSO problem, which is based on the observation that the rLASSO estimate for linear regression parameters can be interpreted as a Bayesian posterior mode estimate when the regression parameters are assigned independent inverse Laplace priors. Bayesian inference from this posterior is possible using an expanded hierarchy motivated by a scale mixture of double Pareto or truncated normal distributions. On simulated and real datasets, we show that the Bayesian formulation outperforms its classical cousin in estimation, prediction, and variable selection across a wide range of scenarios while offering the advantage of posterior inference. Finally, we discuss other variants of this new approach and provide a unified framework for variable selection using flexible reciprocal penalties. All methods described in this article are publicly available as an R package at: https://github.com/himelmallick/BayesRecipe.

Deep Dive into Machine Learning Models for Protein Engineering.

Protein redesign and engineering has become an important task in pharmaceutical research and development. Recent advances in technology have enabled efficient protein redesign by mimicking natural evolutionary mutation, selection, and amplification steps in the laboratory environment. For any given protein, the number of possible mutations is astronomical. It is impractical to synthesize all sequences or even to investigate all functionally interesting variants. Recently, there has been an increased interest in using machine learning to assist protein redesign, since prediction models can be used to virtually screen a large number of novel sequences. However, many state-of-the-art machine learning models, especially deep learning models, have not been extensively explored. Moreover, only a small selection of protein sequence descriptors has been considered. In this work, the performance of prediction models built using an array of machine learning methods and protein descriptor types, including two novel, single amino acid descriptors and one structure-based three-dimensional descriptor, is benchmarked. The predictions were evaluated on a diverse collection of public and proprietary data sets, using a variety of evaluation metrics. The results of this comparison suggest that Convolution Neural Network models built with amino acid property descriptors are the most widely applicable to the types of protein redesign problems faced in the pharmaceutical industry.

Building Quantitative Structure-Activity Relationship Models Using Bayesian Additive Regression Trees.

Quantitative structure-activity relationship (QSAR) is a very commonly used technique for predicting the biological activity of a molecule using information contained in the molecular descriptors. The large number of compounds and descriptors and the sparseness of descriptors pose important challenges to traditional statistical methods and machine learning (ML) algorithms (such as random forest (RF)) used in this field. Recently, Bayesian Additive Regression Trees (BART), a flexible Bayesian nonparametric regression approach, has been demonstrated to be competitive with widely used ML approaches. Instead of only focusing on accurate point estimation, BART is formulated entirely in a hierarchical Bayesian modeling framework, allowing one to also quantify uncertainties and hence to provide both point and interval estimation for a variety of quantities of interest. We studied BART as a model builder for QSAR and demonstrated that the approach tends to have predictive performance comparable to RF. More importantly, we investigated BART's natural capability to analyze truncated (or qualified) data, generate interval estimates for molecular activities as well as descriptor importance, and conduct model diagnosis, which could not be easily handled through other approaches.

EEG spectral analysis of NREM sleep in a large sample of patients with insomnia and good sleepers: effects of age, sex and part of the night.

Previous studies of the differences between patients with insomnia and good sleepers with regard to quantitative electroencephalographic measures have mostly utilized small samples and consequently had limited ability to account for potentially important confounding factors of age, sex and part of the night. We conducted a power spectral analysis using a large database of sleep electroencephalographic recordings to evaluate differences between patients with insomnia (N = 803) and good sleepers (N = 811), while simultaneously accounting for these factors and their interaction. Comparisons of power as a function of age and part of the night were made between cohorts (patients with insomnia versus good sleepers) by sex. Absolute power in the delta, theta and sigma bands declined with age for both females and males. Females had significantly greater power than males at all ages, and for each band, cohort and part of the night. These sex differences were much greater than differences between patients with insomnia and good sleepers. Compared with good sleepers, patients with insomnia under age 40-45 years had reduced delta band power during Part 1 of the night. Females with insomnia over age 45 years had increased delta and theta band power in Parts 2 and 3 of the night, and males with insomnia under age 40 years had reduced theta power in Part 1. Females with insomnia had increased beta2 power in all parts of the night, and males with insomnia had reduced alpha power during all parts of the night. Relative power (the proportion that an individual frequency band contributes to the total power) decreased in the delta band and increased in all other bands with age for both cohorts, sexes and all parts of the night. This analysis provides a unique resource for quantitative information on the differences in power spectra between patients with insomnia and good sleepers accounting for age, sex and part of the night.

Demystifying Multitask Deep Neural Networks for Quantitative Structure-Activity Relationships.

Deep neural networks (DNNs) are complex computational models that have found great success in many artificial intelligence applications, such as computer vision1,2 and natural language processing.3,4 In the past four years, DNNs have also generated promising results for quantitative structure-activity relationship (QSAR) tasks.5,6 Previous work showed that DNNs can routinely make better predictions than traditional methods, such as random forests, on a diverse collection of QSAR data sets. It was also found that multitask DNN models-those trained on and predicting multiple QSAR properties simultaneously-outperform DNNs trained separately on the individual data sets in many, but not all, tasks. To date there has been no satisfactory explanation of why the QSAR of one task embedded in a multitask DNN can borrow information from other unrelated QSAR tasks. Thus, using multitask DNNs in a way that consistently provides a predictive advantage becomes a challenge. In this work, we explored why multitask DNNs make a difference in predictive performance. Our results show that during prediction a multitask DNN does borrow "signal" from molecules with similar structures in the training sets of the other tasks. However, whether this borrowing leads to better or worse predictive performance depends on whether the activities are correlated. On the basis of this, we have developed a strategy to use multitask DNNs that incorporate prior domain knowledge to select training sets with correlated activities, and we demonstrate its effectiveness on several examples.

Efavirenz modulation of sleep spindles and sleep spectral profile.

Non-nucleoside reverse transcriptase inhibitors are important antiretroviral agents for the treatment of human immunodeficiency virus. Some non-nucleoside reverse transcriptase inhibitors, in particular efavirenz, have prominent effects on sleep, cognition and psychiatric variables that limit their tolerability. To avoid confounds due to drug-drug and drug-disease interactions, we assessed the effects of efavirenz in healthy volunteers on sleep, cognition and psychological endpoints during the first week of treatment. Forty healthy male subjects were randomized to receive placebo or efavirenz 600 mg nightly for 7 days after completion of a 3-day placebo run-in period. Treatment with efavirenz was associated with reduced time to sleep onset in the Maintenance of Wakefulness Test, an increase in non-rapid eye movement sleep, a large exposure-related decrease in sigma band spectral density and sleep spindle density during non-rapid eye movement sleep, and reduced performance on an attention switching task. Because efavirenz has been shown to have serotonin 2A receptor partial-agonist properties, we reasoned that antagonism of serotonin 2A receptor signalling in the thalamic reticular nucleus, which generates sleep spindles and promotes attention, may be responsible. Consistent with predictions, treatment of healthy volunteers with a single dose of a serotonin 2A receptor antagonist was found to significantly suppress sigma band spectral density in an exposure-related manner and modulated the overall spectral profile in a manner highly similar to that observed with efavirenz, consistent with the notion that efavirenz exhibits serotonin 2A receptor partial-agonist pharmacology in humans.

Deep neural nets as a method for quantitative structure-activity relationships.

Neural networks were widely used for quantitative structure-activity relationships (QSAR) in the 1990s. Because of various practical issues (e.g., slow on large problems, difficult to train, prone to overfitting, etc.), they were superseded by more robust methods like support vector machine (SVM) and random forest (RF), which arose in the early 2000s. The last 10 years has witnessed a revival of neural networks in the machine learning community thanks to new methods for preventing overfitting, more efficient training algorithms, and advancements in computer hardware. In particular, deep neural nets (DNNs), i.e. neural nets with more than one hidden layer, have found great successes in many applications, such as computer vision and natural language processing. Here we show that DNNs can routinely make better prospective predictions than RF on a set of large diverse QSAR data sets that are taken from Merck's drug discovery effort. The number of adjustable parameters needed for DNNs is fairly large, but our results show that it is not necessary to optimize them for individual data sets, and a single set of recommended parameters can achieve better performance than RF for most of the data sets we studied. The usefulness of the parameters is demonstrated on additional data sets not used in the calibration. Although training DNNs is still computationally intensive, using graphical processing units (GPUs) can make this issue manageable.

A robust bayesian estimate of the concordance correlation coefficient.

A need for assessment of agreement arises in many situations including statistical biomarker qualification or assay or method validation. Concordance correlation coefficient (CCC) is one of the most popular scaled indices reported in evaluation of agreement. Robust methods for CCC estimation currently present an important statistical challenge. Here, we propose a novel Bayesian method of robust estimation of CCC based on multivariate Student's t-distribution and compare it with its alternatives. Furthermore, we extend the method to practically relevant settings, enabling incorporation of confounding covariates and replications. The superiority of the new approach is demonstrated using simulation as well as real datasets from biomarker application in electroencephalography (EEG). This biomarker is relevant in neuroscience for development of treatments for insomnia.

A Bayesian estimate of the concordance correlation coefficient with skewed data.

Concordance correlation coefficient (CCC) is one of the most popular scaled indices used to evaluate agreement. Most commonly, it is used under the assumption that data is normally distributed. This assumption, however, does not apply to skewed data sets. While methods for the estimation of the CCC of skewed data sets have been introduced and studied, the Bayesian approach and its comparison with the previous methods has been lacking. In this study, we propose a Bayesian method for the estimation of the CCC of skewed data sets and compare it with the best method previously investigated. The proposed method has certain advantages. It tends to outperform the best method studied before when the variation of the data is mainly from the random subject effect instead of error. Furthermore, it allows for greater flexibility in application by enabling incorporation of missing data, confounding covariates, and replications, which was not considered previously. The superiority of this new approach is demonstrated using simulation as well as real-life biomarker data sets used in an electroencephalography clinical study. The implementation of the Bayesian method is accessible through the Comprehensive R Archive Network.

A short note on jackknifing the concordance correlation coefficient.

Lin's concordance correlation coefficient (CCC) is a very popular scaled index of agreement used in applied statistics. To obtain a confidence interval (CI) for the estimate of CCC, jackknifing was proposed and shown to perform well in simulation as well as in applications. However, a theoretical proof of the validity of the jackknife CI for the CCC has not been presented yet. In this note, we establish a sufficient condition for using the jackknife method to construct the CI for the CCC.

A comparison of confidence interval methods for the concordance correlation coefficient and intraclass correlation coefficient with small number of raters.

The intraclass correlation coefficient (ICC) with fixed raters or, equivalently, the concordance correlation coefficient (CCC) for continuous outcomes is a widely accepted aggregate index of agreement in settings with small number of raters. Quantifying the precision of the CCC by constructing its confidence interval (CI) is important in early drug development applications, in particular in qualification of biomarker platforms. In recent years, there have been several new methods proposed for construction of CIs for the CCC, but their comprehensive comparison has not been attempted. The methods consisted of the delta method and jackknifing with and without Fisher's Z-transformation, respectively, and Bayesian methods with vague priors. In this study, we carried out a simulation study, with data simulated from multivariate normal as well as heavier tailed distribution (t-distribution with 5 degrees of freedom), to compare the state-of-the-art methods for assigning CI to the CCC. When the data are normally distributed, the jackknifing with Fisher's Z-transformation (JZ) tended to provide superior coverage and the difference between it and the closest competitor, the Bayesian method with the Jeffreys prior was in general minimal. For the nonnormal data, the jackknife methods, especially the JZ method, provided the coverage probabilities closest to the nominal in contrast to the others which yielded overly liberal coverage. Approaches based upon the delta method and Bayesian method with conjugate prior generally provided slightly narrower intervals and larger lower bounds than others, though this was offset by their poor coverage. Finally, we illustrated the utility of the CIs for the CCC in an example of a wake after sleep onset (WASO) biomarker, which is frequently used in clinical sleep studies of drugs for treatment of insomnia.

Development and Evaluation of Conformal Prediction Methods for Quantitative Structure-Activity Relationship.

The quantitative structure-activity relationship (QSAR) regression model is a commonly used technique for predicting the biological activities of compounds using their molecular descriptors. Besides accurate activity estimation, obtaining a prediction uncertainty metric like a prediction interval is highly desirable. Quantifying prediction uncertainty is an active research area in statistical and machine learning (ML), but the implementation for QSAR remains challenging. However, most ML algorithms with high predictive performance require add-on companions for estimating the uncertainty of their prediction. Conformal prediction (CP) is a promising approach as its main components are agnostic to the prediction modes, and it produces valid prediction intervals under weak assumptions on the data distribution. We proposed computationally efficient CP algorithms tailored to the most widely used ML models, including random forests, deep neural networks, and gradient boosting. The algorithms use a novel approach to the derivation of nonconformity scores from the estimates of prediction uncertainty generated by the ensembles of point predictions. The validity and efficiency of proposed algorithms are demonstrated on a diverse collection of QSAR data sets as well as simulation studies. The provided software implementing our algorithms can be used as stand-alone or easily incorporated into other ML software packages for QSAR modeling.

Guidelines for the recording and evaluation of pharmaco-sleep studies in man: the International Pharmaco-EEG Society (IPEG).

The International Pharmaco-EEG Society (IPEG) presents guidelines summarising the requirements for the recording and computerised evaluation of pharmaco-sleep data in man. Over the past years, technical and data-processing methods have advanced steadily, thus enhancing data quality and expanding the palette of sleep assessment tools that can be used to investigate the activity of drugs on the central nervous system (CNS), determine the time course of effects and pharmacodynamic properties of novel therapeutics, hence enabling the study of the pharmacokinetic/pharmacodynamic relationship, and evaluate the CNS penetration or toxicity of compounds. However, despite the presence of robust guidelines on the scoring of polysomnography -recordings, a review of the literature reveals inconsistent -aspects in the operating procedures from one study to another. While this fact does not invalidate results, the lack of standardisation constitutes a regrettable shortcoming, especially in the context of drug development programmes. The present guidelines are intended to assist investigators, who are using pharmaco-sleep measures in clinical research, in an effort to provide clear and concise recommendations and thereby to standardise methodology and facilitate comparability of data across laboratories.

Comparison of polysomnography in people with Alzheimer's disease and insomnia versus non-demented elderly people with insomnia.

BACKGROUND: We used baseline polysomnography (PSG) data obtained during the clinical program development for suvorexant to compare the PSG profiles of people with Alzheimer's disease and insomnia (ADI) versus age-matched elderly individuals with insomnia (EI). METHODS: Sleep laboratory baseline PSG data from participants age 55-80 years from 2 trials in people with insomnia and a trial in people with ADI were included. ADI participants had dementia of mild-to-moderate severity. Diagnostic criteria for insomnia, exclusion for other sleep problems, PSG recording procedures, and endpoint derivations were similar across the trials. All participants underwent a night of in-laboratory PSG prior to the baseline night to allow for screening/adaptation. Participants in the EI and ADI groups were compared with regard to sleep architecture, sleep micro-structure, and quantitative EEG power spectral endpoints. The analysis was performed on a post hoc basis using propensity score matching to compare sleep parameters separately in women and men while accounting for age group and total sleep time. RESULTS: A total of 837 EI and 239 ADI participants were included, with the majority in each population (∼65%) being women. Compared to EI, those with ADI had a lower percentage of time spent in slow wave sleep (and a corresponding higher percentage of time spent in the lighter N1 sleep), a lower number of spindles per minute of N2 sleep, and lower absolute EEG power during NREM sleep, particularly in the lower-frequency bands. Trends for lower REM sleep percentage in ADI did not reach statistical significance. CONCLUSIONS: Our findings in this large data set, in which the influence of sleep problems was effectively subtracted out (since both groups had insomnia), provide strong confirmatory support of results from previous smaller studies in indicating that AD of mild-to-moderate severity is associated with less slow wave sleep, spindles, and lower-frequency EEG power. TRIAL REGISTRATION: ClinicalTrials.gov, numbers NCT01097616, NCT01097629, NCT02750306.

Alterations in cyclic alternating pattern associated with phase advanced sleep are differentially modulated by gaboxadol and zolpidem.

OBJECTIVE: to evaluate cyclic alternating pattern (CAP) in a phase advance model of transient insomnia and the effects of gaboxadol and zolpidem. DESIGN: a randomized, double-blind, cross-over study in which habitual sleep time was advanced by 4 h. SETTING: 6 sleep research laboratories in US PARTICIPANTS: 55 healthy subjects (18-57 y) INTERVENTIONS: Gaboxadol 15 mg (GBX), zolpidem 10 mg (ZOL), and placebo (PBO). MEASUREMENTS: routine polysomnographic (PSG) measures, CAP, spectral power density, and self-reported sleep measures RESULTS: The phase advance model of transient insomnia produced significant changes in CAP parameters. Both GBX and ZOL significantly and differentially modified CAP parameters in the direction of more stable sleep. GBX brought the CAP rate in stage 1 sleep and slow wave sleep (SWS) closer to baseline levels but did not significantly change the CAP rate in stage 2. ZOL reduced the CAP rate in stage 2 to near baseline levels, whereas the CAP rate in stage 1 and SWS was reduced substantially below baseline levels. The CAP parameter A1 index (associated with SWS and sleep continuity) showed the highest correlation with self-reported sleep quality, higher than any traditional PSG, spectral, or other self-reported measures. CONCLUSION: disruptions in CAP produced by phase advanced sleep were significantly and differentially modulated by gaboxadol and zolpidem. The relative independence of CAP parameters from other electrophysiological measures of sleep, their high sensitivity to sleep disruption, and their strong association with subjective sleep quality suggest that CAP variables may serve as valuable endpoints in future insomnia research.

A Deep Learning Approach for Automated Sleep-Wake Scoring in Pre-Clinical Animal Models.

BACKGROUND: Experimental investigation of sleep-wake dynamics in animals is an important part of pharmaceutical development. Typically, it involves recording of electroencephalogram, electromyogram, locomotor activity, and electrooculogram. Visual identification, or scoring, of the sleep-wake states from these recordings is time-consuming. We sought to develop software for automated sleep-wake scoring capable of processing large databases of multi-channel signal recordings in a range of species. NEW METHOD: We used a large historical database of signal recordings and scores in non-human primates, dogs, mice, and rats, to develop a deep Convolutional Neural Network (CNN) classification algorithm for automatically scoring sleep-wake states. We compared the performance of the CNN algorithm with that of a widely used Machine Learning algorithm, Random Forest (RF). RESULTS: CNN accuracy in sleep-wake scoring of data in non-human primates and dogs was significantly higher than RF accuracy (0.75 vs. 0.66 for non-human primates and 0.73 vs. 0.64 for dogs). In rodents, the difference between CNN and RF was smaller: 0.83 vs. 0.81 for mice and 0.78 vs. 0.77 for rats. The variability of CNN accuracy was lower than that of RF for non-human primates, dogs and mice but similar for rats. COMPARISON WITH EXISTING METHODS: Deep Learning algorithms have not been previously evaluated across a range of species for animal sleep-wake scoring. CONCLUSIONS: We recommend use of CNN for sleep-wake scoring in non-human primates and dogs, and RF for sleep-wake scoring in rodents.

How well can a large number of polysomnography sleep measures predict subjective sleep quality in insomnia patients?

OBJECTIVE: The determinants of sleep quality (sQUAL) are poorly understood. We evaluated how well a large number of objective polysomnography (PSG) parameters can predict sQUAL in insomnia patients participating in trials of sleep medications or placebo. METHODS: PSG recordings over multiple nights from two clinical drug development programs involving 1158 insomnia patients treated with suvorexant or placebo and 903 insomnia patients treated with gaboxadol or placebo were used post-hoc to analyze univariate and multivariate associations between sQUAL and 98 PSG sleep parameters plus patient's age and gender. Analyses were performed separately for each of the two clinical trial databases. For univariate associations, within-subject correlations were estimated using mixed effect modeling of bi-variate longitudinal data with one variable being a given PSG variable and the other being sQUAL. To evaluate how accurately sQUAL could be predicted by all PSG variables jointly plus patient's age and gender, the Random Forest multivariate technique was used. Random Forest was also used to evaluate the accuracy of sQUAL prediction by subjective sleep measures plus age and gender, and to quantitatively describe the relative importance of each variable for predicting sQUAL. RESULTS: In the univariate analyses, total sleep time (TST) had the largest correlation with sQUAL compared with all other PSG sleep parameters, and the magnitude of the correlation between each PSG sleep architecture parameter and sQUAL generally increased with the strength of their associations with TST. In the multivariate analyses, the overall accuracy of sQUAL prediction, even with the large number of PSG parameters plus patient's age and gender, was moderate (area under the Receiver Operating Characteristic curve (AROC): 71.2-71.8%). Ranking of PSG parameters by their contribution to sQUAL indicated that TST was the most important predictor of sQUAL among all PSG variables. Subjective TST and subjective number of awakenings jointly with patient's age classified sQUAL with higher accuracy (AROC: 78.7-81.7%) than PSG variables plus age and gender. The pattern of findings was consistent across the two clinical trial databases. CONCLUSION: In insomnia patients participating in trials of sleep medications or placebo, PSG variables had a moderate but consistent pattern of association with sQUAL across two separate clinical trial databases. Of the PSG variables evaluated, TST was the best predictor of sQUAL. CLINICAL TRIALS: trial registration at www.clinicaltrials.gov: NCT01097616; NCT01097629; NCT00094627; NCT00094666.