Digital technologies such as smartphones are transforming the way scientists conduct biomedical research using real-world data. Several remotely-conducted studies have recruited thousands of participants over a span of a few months. Unfortunately, these studies are hampered by substantial participant attrition, calling into question the representativeness of the collected data including generalizability of findings from these studies. We report the challenges in retention and recruitment in eight remote digital health studies comprising over 100,000 participants who participated for more than 850,000 days, completing close to 3.5 million remote health evaluations. Survival modeling surfaced several factors significantly associated(P < 1e-16) with increase in median retention time i) Clinician referral(increase of 40 days), ii) Effect of compensation (22 days), iii) Clinical conditions of interest to the study (7 days) and iv)Older adults(4 days). Additionally, four distinct patterns of daily app usage behavior that were also associated(P < 1e-10) with participant demographics were identified. Most studies were not able to recruit a representative sample, either demographically or regionally. Combined together these findings can help inform recruitment and retention strategies to enable equitable participation of populations in future digital health research.

SUMMARY We present a consensus atlas of the human brain transcriptome in Alzheimer’s disease (AD), based on meta-analysis of differential gene expression in 2,114 postmortem samples. We discover 30 brain coexpression modules from seven regions as the major source of AD transcriptional perturbations. We next examine overlap with 251 brain differentially expressed gene sets from mouse models of AD and other neurodegenerative disorders. Human-mouse overlaps highlight responses to amyloid versus tau pathology and reveal age- and sex-dependent expression signatures for disease progression. Human coexpression modules enriched for neuronal and/or microglial genes broadly overlap with mouse models of AD, Huntington’s disease, amyotrophic lateral sclerosis, and aging. Other human coexpression modules, including those implicated in proteostasis, are not activated in AD models but rather following other, unexpected genetic manipulations. Our results comprise a cross-species resource, highlighting transcriptional networks altered by human brain pathophysiology and identifying correspondences with mouse models for AD preclinical studies.

52Alzheimer's disease (AD) is a complex and heterogenous brain disease that affects multiple inter-related 53 biological processes. This complexity contributes, in part, to existing difficulties in the identification of 54 successful disease-modifying therapeutic strategies. To address this, systems approaches are being used to 55 characterize AD-related disruption in molecular state. To evaluate the consistency across these molecular 56 models, a consensus atlas of the human brain transcriptome was developed through coexpression meta-57 analysis across the AMP-AD consortium. Consensus analysis was performed across five coexpression 58 methods used to analyze RNA-seq data collected from 2114 samples across 7 brain regions and 3 research 59 studies. From this analysis, five consensus clusters were identified that described the major sources of 60 AD-related alterations in transcriptional state that were consistent across studies, methods, and samples. 61AD genetic associations, previously studied AD-related biological processes, and AD targets under active 62 investigation were enriched in only three of these five clusters. The remaining two clusters demonstrated 63 strong heterogeneity between males and females in AD-related expression that was consistently observed 64 across studies. AD transcriptional modules identified by systems analysis of individual AMP-AD teams 65 were all represented in one of these five consensus clusters except ROS/MAP-identified Module 109, 66 which was specific for genes that showed the strongest association with changes in AD-related gene 67 expression across consensus clusters. The other two AMP-AD transcriptional analyses reported modules 68 that were enriched in one of the two sex-specific Consensus Clusters. The fifth cluster has not been 69 previously identified and was enriched for genes related to proteostasis. This study provides an atlas to 70 map across biological inquiries of AD with the goal of supporting an expansion in AD target discovery 71 efforts.

Collection of high-dimensional, longitudinal digital health data has the potential to support a wide-variety of research and clinical applications including diagnostics and longitudinal health tracking. Algorithms that process these data and inform digital diagnostics are typically developed using training and test sets generated from multiple repeated measures collected across a set of individuals. However, the inclusion of repeated measurements is not always appropriately taken into account in the analytical evaluations of predictive performance. The assignment of repeated measurements from each individual to both the training and the test sets (“record-wise” data split) is a common practice and can lead to massive underestimation of the prediction error due to the presence of “identity confounding.” In essence, these models learn to identify subjects, in addition to diagnostic signal. Here, we present a method that can be used to effectively calculate the amount of identity confounding learned by classifiers developed using a record-wise data split. By applying this method to several real datasets, we demonstrate that identity confounding is a serious issue in digital health studies and that record-wise data splits for machine learning- based applications need to be avoided.

Consumer wearables and sensors are a rich source of data about patients’ daily disease and symptom burden, particularly in the case of movement disorders like Parkinson’s disease (PD). However, interpreting these complex data into so-called digital biomarkers requires complicated analytical approaches, and validating these biomarkers requires sufficient data and unbiased evaluation methods. Here we describe the use of crowdsourcing to specifically evaluate and benchmark features derived from accelerometer and gyroscope data in two different datasets to predict the presence of PD and severity of three PD symptoms: tremor, dyskinesia, and bradykinesia. Forty teams from around the world submitted features, and achieved drastically improved predictive performance for PD status (best AUROC = 0.87), as well as tremor- (best AUPR = 0.75), dyskinesia- (best AUPR = 0.48) and bradykinesia-severity (best AUPR = 0.95).