Personalized mood prediction from patterns of behavior collected with smartphones.

Over the last ten years, there has been considerable progress in using digital behavioral phenotypes, captured passively and continuously from smartphones and wearable devices, to infer depressive mood. However, most digital phenotype studies suffer from poor replicability, often fail to detect clinically relevant events, and use measures of depression that are not validated or suitable for collecting large and longitudinal data. Here, we report high-quality longitudinal validated assessments of depressive mood from computerized adaptive testing paired with continuous digital assessments of behavior from smartphone sensors for up to 40 weeks on 183 individuals experiencing mild to severe symptoms of depression. We apply a combination of cubic spline interpolation and idiographic models to generate individualized predictions of future mood from the digital behavioral phenotypes, achieving high prediction accuracy of depression severity up to three weeks in advance (R2 ≥ 80%) and a 65.7% reduction in the prediction error over a baseline model which predicts future mood based on past depression severity alone. Finally, our study verified the feasibility of obtaining high-quality longitudinal assessments of mood from a clinical population and predicting symptom severity weeks in advance using passively collected digital behavioral data. Our results indicate the possibility of expanding the repertoire of patient-specific behavioral measures to enable future psychiatric research.

SLIViT: a general AI framework for clinical-feature diagnosis from limited 3D biomedical-imaging data.

We present SLIViT, a deep-learning framework that accurately measures disease-related risk factors in volumetric biomedical imaging, such as magnetic resonance imaging (MRI) scans, optical coherence tomography (OCT) scans, and ultrasound videos. To evaluate SLIViT, we applied it to five different datasets of these three different data modalities tackling seven learning tasks (including both classification and regression) and found that it consistently and significantly outperforms domain-specific state-of-the-art models, typically improving performance (ROC AUC or correlation) by 0.1-0.4. Notably, compared to existing approaches, SLIViT can be applied even when only a small number of annotated training samples is available, which is often a constraint in medical applications. When trained on less than 700 annotated volumes, SLIViT obtained accuracy comparable to trained clinical specialists while reducing annotation time by a factor of 5,000 demonstrating its utility to automate and expedite ongoing research and other practical clinical scenarios.

Deep learning-based phenotype imputation on population-scale biobank data increases genetic discoveries.

Biobanks that collect deep phenotypic and genomic data across many individuals have emerged as a key resource in human genetics. However, phenotypes in biobanks are often missing across many individuals, limiting their utility. We propose AutoComplete, a deep learning-based imputation method to impute or 'fill-in' missing phenotypes in population-scale biobank datasets. When applied to collections of phenotypes measured across ~300,000 individuals from the UK Biobank, AutoComplete substantially improved imputation accuracy over existing methods. On three traits with notable amounts of missingness, we show that AutoComplete yields imputed phenotypes that are genetically similar to the originally observed phenotypes while increasing the effective sample size by about twofold on average. Further, genome-wide association analyses on the resulting imputed phenotypes led to a substantial increase in the number of associated loci. Our results demonstrate the utility of deep learning-based phenotype imputation to increase power for genetic discoveries in existing biobank datasets.

Phenotype integration improves power and preserves specificity in biobank-based genetic studies of major depressive disorder.

Biobanks often contain several phenotypes relevant to diseases such as major depressive disorder (MDD), with partly distinct genetic architectures. Researchers face complex tradeoffs between shallow (large sample size, low specificity/sensitivity) and deep (small sample size, high specificity/sensitivity) phenotypes, and the optimal choices are often unclear. Here we propose to integrate these phenotypes to combine the benefits of each. We use phenotype imputation to integrate information across hundreds of MDD-relevant phenotypes, which significantly increases genome-wide association study (GWAS) power and polygenic risk score (PRS) prediction accuracy of the deepest available MDD phenotype in UK Biobank, LifetimeMDD. We demonstrate that imputation preserves specificity in its genetic architecture using a novel PRS-based pleiotropy metric. We further find that integration via summary statistics also enhances GWAS power and PRS predictions, but can introduce nonspecific genetic effects depending on input. Our work provides a simple and scalable approach to improve genetic studies in large biobanks by integrating shallow and deep phenotypes.

A biobank-scale test of marginal epistasis reveals genome-wide signals of polygenic epistasis.

The contribution of epistasis (interactions among genes or genetic variants) to human complex trait variation remains poorly understood. Methods that aim to explicitly identify pairs of genetic variants, usually single nucleotide polymorphisms (SNPs), associated with a trait suffer from low power due to the large number of hypotheses tested while also having to deal with the computational problem of searching over a potentially large number of candidate pairs. An alternate approach involves testing whether a single SNP modulates variation in a trait against a polygenic background. While overcoming the limitation of low power, such tests of polygenic or marginal epistasis (ME) are infeasible on Biobank-scale data where hundreds of thousands of individuals are genotyped over millions of SNPs. We present a method to test for ME of a SNP on a trait that is applicable to biobank-scale data. We performed extensive simulations to show that our method provides calibrated tests of ME. We applied our method to test for ME at SNPs that are associated with 53 quantitative traits across ≈ 300 K unrelated white British individuals in the UK Biobank (UKBB). Testing 15, 601 trait-loci associations that were significant in GWAS, we identified 16 trait-loci pairs across 12 traits that demonstrate strong evidence of ME signals (p-value p<5×10-853). We further partitioned the significant ME signals across the genome to identify 6 trait-loci pairs with evidence of local (within-chromosome) ME while 15 show evidence of distal (cross-chromosome) ME. Across the 16 trait-loci pairs, we document that the proportion of trait variance explained by ME is about 12x as large as that explained by the GWAS effects on average (range: 0.59 to 43.89). Our results show, for the first time, evidence of interaction effects between individual genetic variants and overall polygenic background modulating complex trait variation.

The genetic underpinnings of variable penetrance and expressivity of pathogenic mutations in cardiometabolic traits.

Over three percent of people carry a dominant pathogenic mutation, yet only a fraction of carriers develop disease (incomplete penetrance), and phenotypes from mutations in the same gene range from mild to severe (variable expressivity). Here, we investigate underlying mechanisms for this heterogeneity: variable variant effect sizes, carrier polygenic backgrounds, and modulation of carrier effect by genetic background (epistasis). We leveraged exomes and clinical phenotypes from the UK Biobank and the Mt. Sinai Bio Me Biobank to identify carriers of pathogenic variants affecting cardiometabolic traits. We employed recently developed methods to study these cohorts, observing strong statistical support and clinical translational potential for all three mechanisms of variable penetrance and expressivity. For example, scores from our recent model of variant pathogenicity were tightly correlated with phenotype amongst clinical variant carriers, they predicted effects of variants of unknown significance, and they distinguished gain- from loss-of-function variants. We also found that polygenic scores predicted phenotypes amongst pathogenic carriers and that epistatic effects can exceed main carrier effects by an order of magnitude.

Fast kernel-based association testing of non-linear genetic effects for biobank-scale data.

Our knowledge of non-linear genetic effects on complex traits remains limited, in part, due to the modest power to detect such effects. While kernel-based tests offer a versatile approach to test for non-linear relationships between sets of genetic variants and traits, current approaches cannot be applied to Biobank-scale datasets containing hundreds of thousands of individuals. We propose, FastKAST, a kernel-based approach that can test for non-linear effects of a set of variants on a quantitative trait. FastKAST provides calibrated hypothesis tests while enabling analysis of Biobank-scale datasets with hundreds of thousands of unrelated individuals from a homogeneous population. We apply FastKAST to 53 quantitative traits measured across ≈ 300 K unrelated white British individuals in the UK Biobank to detect sets of variants with non-linear effects at genome-wide significance.

Tractable and Expressive Generative Models of Genetic Variation Data.

Population genetic studies often rely on artificial genomes (AGs) simulated by generative models of genetic data. In recent years, unsupervised learning models, based on hidden Markov models, deep generative adversarial networks, restricted Boltzmann machines, and variational autoencoders, have gained popularity due to their ability to generate AGs closely resembling empirical data. These models, however, present a tradeoff between expressivity and tractability. Here, we propose to use hidden Chow-Liu trees (HCLTs) and their representation as probabilistic circuits (PCs) as a solution to this tradeoff. We first learn an HCLT structure that captures the long-range dependencies among SNPs in the training data set. We then convert the HCLT to its equivalent PC as a means of supporting tractable and efficient probabilistic inference. The parameters in these PCs are inferred with an expectation-maximization algorithm using the training data. Compared to other models for generating AGs, HCLT obtains the largest log-likelihood on test genomes across SNPs chosen across the genome and from a contiguous genomic region. Moreover, the AGs generated by HCLT more accurately resemble the source data set in their patterns of allele frequencies, linkage disequilibrium, pairwise haplotype distances, and population structure. This work not only presents a new and robust AG simulator but also manifests the potential of PCs in population genetics.

Leveraging family data to design Mendelian randomization that is provably robust to population stratification.

Mendelian randomization (MR) has emerged as a powerful approach to leverage genetic instruments to infer causality between pairs of traits in observational studies. However, the results of such studies are susceptible to biases owing to weak instruments, as well as the confounding effects of population stratification and horizontal pleiotropy. Here, we show that family data can be leveraged to design MR tests that are provably robust to confounding from population stratification, assortative mating, and dynastic effects. We show in simulations that our approach, MR-Twin, is robust to confounding from population stratification and is not affected by weak instrument bias, whereas standard MR methods yield inflated false positive rates. We then conduct an exploratory analysis of MR-Twin and other MR methods applied to 121 trait pairs in the UK Biobank data set. Our results suggest that confounding from population stratification can lead to false positives for existing MR methods, whereas MR-Twin is immune to this type of confounding, and that MR-Twin can help assess whether traditional approaches may be inflated owing to confounding from population stratification.

Causal effects on complex traits are similar for common variants across segments of different continental ancestries within admixed individuals.

Individuals of admixed ancestries (for example, African Americans) inherit a mosaic of ancestry segments (local ancestry) originating from multiple continental ancestral populations. This offers the unique opportunity of investigating the similarity of genetic effects on traits across ancestries within the same population. Here we introduce an approach to estimate correlation of causal genetic effects (radmix) across local ancestries and analyze 38 complex traits in African-European admixed individuals (N = 53,001) to observe very high correlations (meta-analysis radmix = 0.95, 95% credible interval 0.93-0.97), much higher than correlation of causal effects across continental ancestries. We replicate our results using regression-based methods from marginal genome-wide association study summary statistics. We also report realistic scenarios where regression-based methods yield inflated heterogeneity-by-ancestry due to ancestry-specific tagging of causal effects, and/or polygenicity. Our results motivate genetic analyses that assume minimal heterogeneity in causal effects by ancestry, with implications for the inclusion of ancestry-diverse individuals in studies.

The lingering effects of Neanderthal introgression on human complex traits.

The genetic variants introduced into the ancestors of modern humans from interbreeding with Neanderthals have been suggested to contribute an unexpected extent to complex human traits. However, testing this hypothesis has been challenging due to the idiosyncratic population genetic properties of introgressed variants. We developed rigorous methods to assess the contribution of introgressed Neanderthal variants to heritable trait variation and applied these methods to analyze 235,592 introgressed Neanderthal variants and 96 distinct phenotypes measured in about 300,000 unrelated white British individuals in the UK Biobank. Introgressed Neanderthal variants make a significant contribution to trait variation (explaining 0.12% of trait variation on average). However, the contribution of introgressed variants tends to be significantly depleted relative to modern human variants matched for allele frequency and linkage disequilibrium (about 59% depletion on average), consistent with purifying selection on introgressed variants. Different from previous studies (McArthur et al., 2021), we find no evidence for elevated heritability across the phenotypes examined. We identified 348 independent significant associations of introgressed Neanderthal variants with 64 phenotypes. Previous work (Skov et al., 2020) has suggested that a majority of such associations are likely driven by statistical association with nearby modern human variants that are the true causal variants. Applying a customized fine-mapping led us to identify 112 regions across 47 phenotypes containing 4303 unique genetic variants where introgressed variants are highly likely to have a phenotypic effect. Examination of these variants reveals their substantial impact on genes that are important for the immune system, development, and metabolism.

MaLAdapt Reveals Novel Targets of Adaptive Introgression From Neanderthals and Denisovans in Worldwide Human Populations.

Adaptive introgression (AI) facilitates local adaptation in a wide range of species. Many state-of-the-art methods detect AI with ad-hoc approaches that identify summary statistic outliers or intersect scans for positive selection with scans for introgressed genomic regions. Although widely used, approaches intersecting outliers are vulnerable to a high false-negative rate as the power of different methods varies, especially for complex introgression events. Moreover, population genetic processes unrelated to AI, such as background selection or heterosis, may create similar genomic signals to AI, compromising the reliability of methods that rely on neutral null distributions. In recent years, machine learning (ML) methods have been increasingly applied to population genetic questions. Here, we present a ML-based method called MaLAdapt for identifying AI loci from genome-wide sequencing data. Using an Extra-Trees Classifier algorithm, our method combines information from a large number of biologically meaningful summary statistics to capture a powerful composite signature of AI across the genome. In contrast to existing methods, MaLAdapt is especially well-powered to detect AI with mild beneficial effects, including selection on standing archaic variation, and is robust to non-AI selective sweeps, heterosis from deleterious mutations, and demographic misspecification. Furthermore, MaLAdapt outperforms existing methods for detecting AI based on the analysis of simulated data and the validation of empirical signals through visual inspection of haplotype patterns. We apply MaLAdapt to the 1000 Genomes Project human genomic data and discover novel AI candidate regions in non-African populations, including genes that are enriched in functionally important biological pathways regulating metabolism and immune responses.

Phenotypic subtyping via contrastive learning.

Defining and accounting for subphenotypic structure has the potential to increase statistical power and provide a deeper understanding of the heterogeneity in the molecular basis of complex disease. Existing phenotype subtyping methods primarily rely on clinically observed heterogeneity or metadata clustering. However, they generally tend to capture the dominant sources of variation in the data, which often originate from variation that is not descriptive of the mechanistic heterogeneity of the phenotype of interest; in fact, such dominant sources of variation, such as population structure or technical variation, are, in general, expected to be independent of subphenotypic structure. We instead aim to find a subspace with signal that is unique to a group of samples for which we believe that subphenotypic variation exists (e.g., cases of a disease). To that end, we introduce Phenotype Aware Components Analysis (PACA), a contrastive learning approach leveraging canonical correlation analysis to robustly capture weak sources of subphenotypic variation. In the context of disease, PACA learns a gradient of variation unique to cases in a given dataset, while leveraging control samples for accounting for variation and imbalances of biological and technical confounders between cases and controls. We evaluated PACA using an extensive simulation study, as well as on various subtyping tasks using genotypes, transcriptomics, and DNA methylation data. Our results provide multiple strong evidence that PACA allows us to robustly capture weak unknown variation of interest while being calibrated and well-powered, far superseding the performance of alternative methods. This renders PACA as a state-of-the-art tool for defining de novo subtypes that are more likely to reflect molecular heterogeneity, especially in challenging cases where the phenotypic heterogeneity may be masked by a myriad of strong unrelated effects in the data.

Leveraging family data to design Mendelian Randomization that is provably robust to population stratification.

Mendelian Randomization (MR) has emerged as a powerful approach to leverage genetic instruments to infer causality between pairs of traits in observational studies. However, the results of such studies are susceptible to biases due to weak instruments as well as the confounding effects of population stratification and horizontal pleiotropy. Here, we show that family data can be leveraged to design MR tests that are provably robust to confounding from population stratification, assortative mating, and dynastic effects. We demonstrate in simulations that our approach, MR-Twin, is robust to confounding from population stratification and is not affected by weak instrument bias, while standard MR methods yield inflated false positive rates. We applied MR-Twin to 121 trait pairs in the UK Biobank dataset and found that MR-Twin identifies likely causal trait pairs and does not identify trait pairs that are unlikely to be causal. Our results suggest that confounding from population stratification can lead to false positives for existing MR methods, while MR-Twin is immune to this type of confounding.

Automated Identification of Incomplete and Complete Retinal Epithelial Pigment and Outer Retinal Atrophy Using Machine Learning.

OBJECTIVE: To assess and validate a deep learning algorithm to automatically detect incomplete retinal pigment epithelial and outer retinal atrophy (iRORA) and complete retinal pigment epithelial and outer retinal atrophy (cRORA) in eyes with age-related macular degeneration. DESIGN: In a retrospective machine learning analysis, a deep learning model was trained to jointly classify the presence of iRORA and cRORA within a given B-scan. The algorithm was evaluated using 2 separate and independent datasets. PARTICIPANTS: OCT B-scan volumes from 71 patients with nonneovascular age-related macular degeneration captured at the Doheny-University of California Los Angeles Eye Centers and the following 2 external OCT B-scans testing datasets: (1) University of Pennsylvania, University of Miami, and Case Western Reserve University and (2) Doheny Image Reading Research Laboratory. METHODS: The images were annotated by an experienced grader for the presence of iRORA and cRORA. A Resnet18 model was trained to classify these annotations for each B-scan using OCT volumes collected at the Doheny-University of California Los Angeles Eye Centers. The model was applied to 2 testing datasets to assess out-of-sample model performance. MAIN OUTCOMES MEASURES: Model performance was quantified in terms of area under the receiver operating characteristic curve (AUROC) and area under the precision-recall curve (AUPRC). Sensitivity, specificity, and positive predictive value were also compared against additional clinician annotators. RESULTS: On an independently collected test set, consisting of 1117 volumes from the general population, the model predicted iRORA and cRORA presence within the entire volume with nearly perfect AUROC performance and AUPRC scores (iRORA, 0.61; 95% confidence interval [CI] [0.45, 0.82]: cRORA, 0.83; 95% CI [0.68, 0.95]). On another independently collected set, consisting of 60 OCT B-scans enriched for iRORA and cRORA lesions, the model performed with AUROC (iRORA: 0.68, 95% CI [0.54, 0.81]; cRORA: 0.84, 95% CI [0.75, 0.94]) and AUPRC (iRORA: 0.70, 95% CI [0.55, 0.86]; cRORA: 0.82, 95% CI [0.70, 0.93]). CONCLUSIONS: A deep learning model can accurately and precisely identify both iRORA and cRORA lesions within the OCT B-scan volume. The model can achieve similar sensitivity compared with human graders, which potentially obviates a laborious and time-consuming annotation process and could be developed into a diagnostic screening tool.

A scalable and robust variance components method reveals insights into the architecture of gene-environment interactions underlying complex traits.

Understanding the contribution of gene-environment interactions (GxE) to complex trait variation can provide insights into mechanisms underlying disease risk, explain sources of heritability, and improve the accuracy of genetic risk prediction. While biobanks that collect genetic and deep phenotypic data over large numbers of individuals offer the promise of obtaining novel insights into GxE, our understanding of the architecture of GxE in complex traits remains limited. We introduce a method that can estimate the proportion of trait variance explained by GxE (GxE heritability) and additive genetic effects (additive heritability) across the genome and within specific genomic annotations. We show that our method is accurate in simulations and computationally efficient for biobank-scale datasets. We applied our method to ≈ 500, 000 common array SNPs (MAF ≥ 1%), fifty quantitative traits, and four environmental variables (smoking, sex, age, and statin usage) measured across ≈ 300, 000 unrelated white British individuals in the UK Biobank. We found 69 trait-environmental variable pairs with significant genome-wide GxE heritability (p < 0.05/200 correcting for the number of trait-E pairs tested) with an average ratio of GxE to additive heritability ≈ 6.8% that include BMI with smoking (ratio of GxE to additive heritability = 6.3 ± 1.1%), WHR (waist-to-hip ratio adjusted for BMI) with sex (ratio = 19.6 ± 2%), LDL cholesterol with age (ratio = 9.8 ± 3.9%), and HbA1c with statin usage (ratio = 11 ± 2%). Analyzing nearly 8 million common and low-frequency imputed SNPs (MAF ≥ 0.1%), we document an increase in genome-wide GxE heritability of about 28% on average over array SNPs. We partitioned GxE heritability across minor allele frequency (MAF) and local linkage disequilibrium values (LD score) of each SNP to observe that analogous to the relationship that has been observed for additive allelic effects, the magnitude of GxE allelic effects tends to increase with decreasing MAF and LD. Testing whether GxE heritability is enriched around genes that are highly expressed in specific tissues, we find significant tissue-specific enrichments that include brain-specific enrichment for BMI and Basal Metabolic Rate in the context of smoking, adipose-specific enrichment for WHR in the context of sex, and cardiovascular tissue-specific enrichment for total cholesterol in the context of age. Our analyses provide detailed insights into the architecture of GxE underlying complex traits.

Cross-trait assortative mating is widespread and inflates genetic correlation estimates.

The observation of genetic correlations between disparate human traits has been interpreted as evidence of widespread pleiotropy. Here, we introduce cross-trait assortative mating (xAM) as an alternative explanation. We observe that xAM affects many phenotypes and that phenotypic cross-mate correlation estimates are strongly associated with genetic correlation estimates (R2=74%). We demonstrate that existing xAM plausibly accounts for substantial fractions of genetic correlation estimates and that previously reported genetic correlation estimates between some pairs of psychiatric disorders are congruent with xAM alone. Finally, we provide evidence for a history of xAM at the genetic level using cross-trait even/odd chromosome polygenic score correlations. Together, our results demonstrate that previous reports have likely overestimated the true genetic similarity between many phenotypes.

Autonomous wearable sweat rate monitoring based on digitized microbubble detection.

Advancements in wearable bioanalytical microsystems have enabled diurnal and (semi)continuous monitoring of physiologically-relevant indices that are accessible through probing sweat. To deliver an undistorted and physiologically-meaningful interpretation of these readings, tracking the sweat secretion rate is essential, because it allows for calibrating the biomarker readings against variations in sweat secretion and inferring the body's hydration/electrolyte homeostasis status. To realize an autonomous wearable solution with intrinsically high signal-to-noise ratio sweat rate sensing capabilities, here, we devise a digitized microbubble detection mechanism-delivered by a hybrid microfluidic/electronic system with a compact footprint. This mechanism is based on the intermittent generation of microliter-scale bubbles via electrolysis and the instantaneous measurement of their time-of-flight (and thus, velocity) via impedimetric sensing. In this way, we overcome the limitations of previously proposed sweat rate sensing modalities that are inherently susceptible to non-targeted secretion characteristics (pH, conductivity, and temperature), constrained by volume, or lack system integration for autonomous on-body operation. By deploying our solution in human subject trials, we validate the utility of our solution for seamless monitoring of exercise- and iontophoretically-induced sweat secretion profiles.

Accurate modeling of replication rates in genome-wide association studies by accounting for Winner's Curse and study-specific heterogeneity.

Genome-wide association studies (GWAS) have identified thousands of genetic variants associated with complex human traits, but only a fraction of variants identified in discovery studies achieve significance in replication studies. Replication in genome-wide association studies has been well-studied in the context of Winner's Curse, which is the inflation of effect size estimates for significant variants due to statistical chance. However, Winner's Curse is often not sufficient to explain lack of replication. Another reason why studies fail to replicate is that there are fundamental differences between the discovery and replication studies. A confounding factor can create the appearance of a significant finding while actually being an artifact that will not replicate in future studies. We propose a statistical framework that utilizes genome-wide association studies and replication studies to jointly model Winner's Curse and study-specific heterogeneity due to confounding factors. We apply this framework to 100 genome-wide association studies from the Human Genome-Wide Association Studies Catalog and observe that there is a large range in the level of estimated confounding. We demonstrate how this framework can be used to distinguish when studies fail to replicate due to statistical noise and when they fail due to confounding.