Pulmonary Fibrosis Stakeholder Summit: A Joint NHLBI, Three Lakes Foundation, and Pulmonary Fibrosis Foundation Workshop Report.

Despite progress in elucidation of disease mechanisms, identification of risk factors, biomarker discovery, and the approval of two medications to slow lung function decline in idiopathic pulmonary fibrosis and one medication to slow lung function decline in progressive pulmonary fibrosis, pulmonary fibrosis remains a disease with a high morbidity and mortality. In recognition of the need to catalyze ongoing advances and collaboration in the field of pulmonary fibrosis, the NHLBI, the Three Lakes Foundation, and the Pulmonary Fibrosis Foundation hosted the Pulmonary Fibrosis Stakeholder Summit on November 8-9, 2022. This workshop was held virtually and was organized into three topic areas: 1) novel models and research tools to better study pulmonary fibrosis and uncover new therapies, 2) early disease risk factors and methods to improve diagnosis, and 3) innovative approaches toward clinical trial design for pulmonary fibrosis. In this workshop report, we summarize the content of the presentations and discussions, enumerating research opportunities for advancing our understanding of the pathogenesis, treatment, and outcomes of pulmonary fibrosis.

Universal risk phenotype of US counties for flu-like transmission to improve county-specific COVID-19 incidence forecasts.

The spread of a communicable disease is a complex spatio-temporal process shaped by the specific transmission mechanism, and diverse factors including the behavior, socio-economic and demographic properties of the host population. While the key factors shaping transmission of influenza and COVID-19 are beginning to be broadly understood, making precise forecasts on case count and mortality is still difficult. In this study we introduce the concept of a universal geospatial risk phenotype of individual US counties facilitating flu-like transmission mechanisms. We call this the Universal Influenza-like Transmission (UnIT) score, which is computed as an information-theoretic divergence of the local incidence time series from an high-risk process of epidemic initiation, inferred from almost a decade of flu season incidence data gleaned from the diagnostic history of nearly a third of the US population. Despite being computed from the past seasonal flu incidence records, the UnIT score emerges as the dominant factor explaining incidence trends for the COVID-19 pandemic over putative demographic and socio-economic factors. The predictive ability of the UnIT score is further demonstrated via county-specific weekly case count forecasts which consistently outperform the state of the art models throughout the time-line of the COVID-19 pandemic. This study demonstrates that knowledge of past epidemics may be used to chart the course of future ones, if transmission mechanisms are broadly similar, despite distinct disease processes and causative pathogens.

Inverse Gillespie for inferring stochastic reaction mechanisms from intermittent samples.

Gillespie stochastic simulation is used extensively to investigate stochastic phenomena in many fields, ranging from chemistry to biology to ecology. The inverse problem, however, has remained largely unsolved: How to reconstruct the underlying reactions de novo from sparse observations. A key challenge is that often only aggregate concentrations, proportional to the population numbers, are observable intermittently. We discovered that under specific assumptions, the set of relative population updates in phase space forms a convex polytope whose vertices are indicative of the dominant underlying reactions. We demonstrate the validity of this simple principle by reconstructing stochastic models (reaction structure plus propensities) from a variety of simulated and experimental systems, where hundreds and even thousands of reactions may be occurring in between observations. In some cases, the inferred models provide mechanistic insight. This principle can lead to the understanding of a broad range of phenomena, from molecular biology to population ecology.

A digital twin of the infant microbiome to predict neurodevelopmental deficits.

Despite the recognized gut-brain axis link, natural variations in microbial profiles between patients hinder definition of normal abundance ranges, confounding the impact of dysbiosis on infant neurodevelopment. We infer a digital twin of the infant microbiome, forecasting ecosystem trajectories from a few initial observations. Using 16S ribosomal RNA profiles from 88 preterm infants (398 fecal samples and 32,942 abundance estimates for 91 microbial classes), the model (Q-net) predicts abundance dynamics with R2 = 0.69. Contrasting the fit to Q-nets of typical versus suboptimal development, we can reliably estimate individual deficit risk (Mδ) and identify infants achieving poor future head circumference growth with ≈76% area under the receiver operator characteristic curve, 95% ± 1.8% positive predictive value at 98% specificity at 30 weeks postmenstrual age. We find that early transplantation might mitigate risk for ≈45.2% of the cohort, with potentially negative effects from incorrect supplementation. Q-nets are generative artificial intelligence models for ecosystem dynamics, with broad potential applications.

Predicting Malignant Ventricular Arrhythmias Using Real-Time Remote Monitoring.

BACKGROUND: Although implantable cardioverter-defibrillator (ICD) therapies are associated with increased morbidity and mortality, the prediction of malignant ventricular arrhythmias has remained elusive. OBJECTIVES: The purpose of this study was to evaluate whether daily remote-monitoring data may predict appropriate ICD therapies for ventricular tachycardia or ventricular fibrillation. METHODS: This was a post hoc analysis of IMPACT (Randomized trial of atrial arrhythmia monitoring to guide anticoagulation in patients with implanted defibrillator and cardiac resynchronization devices), a multicenter, randomized, controlled trial of 2,718 patients evaluating atrial tachyarrhythmias and anticoagulation for patients with heart failure and ICD or cardiac resynchronization therapy with defibrillator devices. All device therapies were adjudicated as either appropriate (to treat ventricular tachycardia or ventricular fibrillation) or inappropriate (all others). Remote monitoring data in the 30 days before device therapy were utilized to develop separate multivariable logistic regression and neural network models to predict appropriate device therapies. RESULTS: A total of 59,807 device transmissions were available for 2,413 patients (age 64 ± 11 years, 26% women, 64% ICD). Appropriate device therapies (141 shocks, 10 antitachycardia pacing) were delivered to 151 patients. Logistic regression identified shock lead impedance and ventricular ectopy as significantly associated with increased risk of appropriate device therapy (sensitivity 39%, specificity 91%, AUC: 0.72). Neural network modeling yielded significantly better (P < 0.01 for comparison) predictive performance (sensitivity 54%, specificity 96%, AUC: 0.90), and also identified patterns of change in atrial lead impedance, mean heart rate, and patient activity as predictors of appropriate therapies. CONCLUSIONS: Daily remote monitoring data may be utilized to predict malignant ventricular arrhythmias in the 30 days before device therapies. Neural networks complement and enhance conventional approaches to risk stratification.

Clinical characterization and differentiation of B-SNIP psychosis Biotypes: Algorithmic Diagnostics for Efficient Prescription of Treatments (ADEPT)-1.

Clinically defined psychosis diagnoses are neurobiologically heterogeneous. The B-SNIP consortium identified and validated more neurobiologically homogeneous psychosis Biotypes using an extensive battery of neurocognitive and psychophysiological laboratory measures. However, typically the first step in any diagnostic evaluation is the clinical interview. In this project, we evaluated if psychosis Biotypes have clinical characteristics that can support their differentiation in addition to obtaining laboratory testing. Clinical interview data from 1907 individuals with a psychosis Biotype were used to create a diagnostic algorithm. The features were 58 ratings from standard clinical scales. Extremely randomized tree algorithms were used to evaluate sensitivity, specificity, and overall classification success. Biotype classification accuracy peaked at 91 % with the use of 57 items on average. A reduced feature set of 28 items, though, also showed 81 % classification accuracy. Using this reduced item set, we found that only 10-11 items achieved a one-vs-all (Biotype-1 or not, Biotype-2 or not, Biotype-3 or not) area under the sensitivity-specificity curve of .78 to .81. The top clinical characteristics for differentiating psychosis Biotypes, in order of importance, were (i) difficulty in abstract thinking, (ii) multiple indicators of social functioning, (iii) conceptual disorganization, (iv) severity of hallucinations, (v) stereotyped thinking, (vi) suspiciousness, (vii) unusual thought content, (viii) lack of spontaneous speech, and (ix) severity of delusions. These features were remarkably different from those that differentiated DSM psychosis diagnoses. This low-burden adaptive algorithm achieved reasonable classification accuracy and will support Biotype-specific etiological and treatment investigations even in under-resourced clinical and research environments.

Event-level prediction of urban crime reveals a signature of enforcement bias in US cities.

Policing efforts to thwart crime typically rely on criminal infraction reports, which implicitly manifest a complex relationship between crime, policing and society. As a result, crime prediction and predictive policing have stirred controversy, with the latest artificial intelligence-based algorithms producing limited insight into the social system of crime. Here we show that, while predictive models may enhance state power through criminal surveillance, they also enable surveillance of the state by tracing systemic biases in crime enforcement. We introduce a stochastic inference algorithm that forecasts crime by learning spatio-temporal dependencies from event reports, with a mean area under the receiver operating characteristic curve of ~90% in Chicago for crimes predicted per week within ~1,000 ft. Such predictions enable us to study perturbations of crime patterns that suggest that the response to increased crime is biased by neighbourhood socio-economic status, draining policy resources from socio-economically disadvantaged areas, as demonstrated in eight major US cities.

Cardiac Comorbidity Risk Score: Zero-Burden Machine Learning to Improve Prediction of Postoperative Major Adverse Cardiac Events in Hip and Knee Arthroplasty.

Background In this retrospective, observational study we introduce the Cardiac Comorbidity Risk Score, predicting perioperative major adverse cardiac events (MACE) after elective hip and knee arthroplasty. MACE is a rare but important driver of mortality, and existing tools, eg, the Revised Cardiac Risk Index demonstrate only modest accuracy. We demonstrate an artificial intelligence-based approach to identify patients at high risk of MACE within 4 weeks (primary outcome) of arthroplasty, that imposes zero additional burden of cost/resources. Methods and Results Cardiac Comorbidity Risk Score calculation uses novel machine learning to estimate MACE risk from patient electronic health records, without requiring blood work or access to any demographic data beyond that of sex and age, and accounts for variable/missing/incomplete information across patient records. Validated on a deidentified cohort (age >45 years, n=445 391), performance was evaluated using the area under the receiver operator characteristics curve (AUROC), sensitivity/specificity, positive predictive value, and positive/negative likelihood ratios. In our cohort (age 63.5±10.5 years, 58.2% women, 34.2%/65.8% hip/knee procedures), 0.19% (882) experienced the primary outcome. Cardiac Comorbidity Risk Score achieved area under the receiver operator characteristics curve=80.0±0.4% (95% CI) for women and 80.1±0.5% (95% CI) for males, with 36.4% and 35.1% sensitivities, respectively, at 95% specificity, significantly outperforming Revised Cardiac Risk Index across all studied age-, sex-, risk-, and comorbidity-based subgroups. Conclusions Cardiac Comorbidity Risk Score, a novel artificial intelligence-based screening tool using known and unknown comorbidity patterns, outperforms state-of-the-art in predicting MACE within 4 weeks postarthroplasty, and can identify patients at high risk that do not demonstrate traditional risk factors.

Development of a computerized adaptive diagnostic screening tool for psychosis.

We develop a two-stage diagnostic classification system for psychotic disorders using an extremely randomized trees machine learning algorithm. Item bank was developed from clinician-rated items drawn from an inpatient and outpatient sample. In stage 1, we differentiate schizophrenia and schizoaffective disorder from depression and bipolar disorder (with psychosis). In stage 2 we differentiate schizophrenia from schizoaffective disorder. Out of sample classification accuracy, determined by area under the receiver operator characteristic (ROC) curve, was outstanding for stage 1 (Area under the ROC curve (AUC) = 0.93, 95% confidence interval (CI) = 0.89, 0.94), and excellent for stage 2 (AUC = 0.86, 95% CI = 0.83, 0.88). This is achieved based on an average of 5 items for stage 1 and an average of 6 items for stage 2, out of a bank of 73 previously validated items.

Screening for idiopathic pulmonary fibrosis using comorbidity signatures in electronic health records.

Idiopathic pulmonary fibrosis (IPF) is a lethal fibrosing interstitial lung disease with a mean survival time of less than 5 years. Nonspecific presentation, a lack of effective early screening tools, unclear pathobiology of early-stage IPF and the need for invasive and expensive procedures for diagnostic confirmation hinder early diagnosis. In this study, we introduce a new screening tool for IPF in primary care settings that requires no new laboratory tests and does not require recognition of early symptoms. Using subtle comorbidity signatures identified from the history of medical encounters of individuals, we developed an algorithm, called the zero-burden comorbidity risk score for IPF (ZCoR-IPF), to predict the future risk of an IPF diagnosis. ZCoR-IPF was trained on a national insurance claims database and validated on three independent databases, comprising a total of 2,983,215 participants, with 54,247 positive cases. The algorithm achieved positive likelihood ratios greater than 30 at a specificity of 0.99 across different cohorts, for both sexes, and for participants with different risk states and history of confounding diseases. The area under the receiver-operating characteristic curve for ZCoR-IPF in predicting IPF exceeded 0.88 and was approximately 0.84 at 1 and 4 years before a conventional diagnosis, respectively. Thus, if adopted, ZCoR-IPF can potentially enable earlier diagnosis of IPF and improve outcomes of disease-modifying therapies and other interventions.

Reduced false positives in autism screening via digital biomarkers inferred from deep comorbidity patterns.

Here, we develop digital biomarkers for autism spectrum disorder (ASD), computed from patterns of past medical encounters, identifying children at high risk with an area under the receiver operating characteristic exceeding 80% from shortly after 2 years of age for either sex, and across two independent patient databases. We leverage uncharted ASD comorbidities, with no requirement of additional blood work, or procedures, to estimate the autism comorbid risk score (ACoR), during the earliest years when interventions are the most effective. ACoR has superior predictive performance to common questionnaire-based screenings and can reduce their current socioeconomic, ethnic, and demographic biases. In addition, we can condition on current screening scores to either halve the state-of-the-art false-positive rate or boost sensitivity to over 60%, while maintaining specificity above 95%. Thus, ACoR can significantly reduce the median diagnostic age, reducing diagnostic delays and accelerating access to evidence-based interventions.

Development and Validation of Computerized Adaptive Assessment Tools for the Measurement of Posttraumatic Stress Disorder Among US Military Veterans.

Importance: Veterans from recent and past conflicts have high rates of posttraumatic stress disorder (PTSD). Adaptive testing strategies can increase accuracy of diagnostic screening and symptom severity measurement while decreasing patient and clinician burden. Objective: To develop and validate a computerized adaptive diagnostic (CAD) screener and computerized adaptive test (CAT) for PTSD symptom severity. Design, Setting, and Participants: A diagnostic study of measure development and validation was conducted at a Veterans Health Administration facility. A total of 713 US military veterans were included. The study was conducted from April 25, 2017, to November 10, 2019. Main Outcomes and Measures: The participants completed a PTSD-symptom questionnaire from the item bank and provided responses on the PTSD Checklist for Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) (PCL-5). A subsample of 304 participants were interviewed using the Clinician-Administered Scale for PTSD for DSM-5. Results: Of the 713 participants, 585 were men; mean (SD) age was 52.8 (15.0) years. The CAD-PTSD reproduced the Clinician-Administered Scale for PTSD for DSM-5 PTSD diagnosis with high sensitivity and specificity as evidenced by an area under the curve of 0.91 (95% CI, 0.87-0.95). The CAT-PTSD demonstrated convergent validity with the PCL-5 (r = 0.88) and also tracked PTSD diagnosis (area under the curve = 0.85; 95% CI, 0.79-0.89). The CAT-PTSD reproduced the final 203-item bank score with a correlation of r = 0.95 with a mean of only 10 adaptively administered items, a 95% reduction in patient burden. Conclusions and Relevance: Using a maximum of only 6 items, the CAD-PTSD developed in this study was shown to have excellent diagnostic screening accuracy. Similarly, using a mean of 10 items, the CAT-PTSD provided valid severity ratings with excellent convergent validity with an extant scale containing twice the number of items. The 10-item CAT-PTSD also outperformed the 20-item PCL-5 in terms of diagnostic accuracy. The results suggest that scalable, valid, and rapid PTSD diagnostic screening and severity measurement are possible.

Postnatal immune activation causes social deficits in a mouse model of tuberous sclerosis: Role of microglia and clinical implications.

There is growing evidence that prenatal immune activation contributes to neuropsychiatric disorders. Here, we show that early postnatal immune activation resulted in profound impairments in social behavior, including in social memory in adult male mice heterozygous for a gene responsible for tuberous sclerosis complex (Tsc2+/−), a genetic disorder with high prevalence of autism. Early postnatal immune activation did not affect either wild-type or female Tsc2+/− mice. We demonstrate that these memory deficits are caused by abnormal mammalian target of rapamycin­dependent interferon signaling and impairments in microglia function. By mining the medical records of more than 3 million children followed from birth, we show that the prevalence of hospitalizations due to infections in males (but not in females) is associated with future development of autism spectrum disorders (ASD). Together, our results suggest the importance of synergistic interactions between strong early postnatal immune activation and mutations associated with ASD.

Estimating heritability and genetic correlations from large health datasets in the absence of genetic data.

Typically, estimating genetic parameters, such as disease heritability and between-disease genetic correlations, demands large datasets containing all relevant phenotypic measures and detailed knowledge of family relationships or, alternatively, genotypic and phenotypic data for numerous unrelated individuals. Here, we suggest an alternative, efficient estimation approach through the construction of two disease metrics from large health datasets: temporal disease prevalence curves and low-dimensional disease embeddings. We present eleven thousand heritability estimates corresponding to five study types: twins, traditional family studies, health records-based family studies, single nucleotide polymorphisms, and polygenic risk scores. We also compute over six hundred thousand estimates of genetic, environmental and phenotypic correlations. Furthermore, we find that: (1) disease curve shapes cluster into five general patterns; (2) early-onset diseases tend to have lower prevalence than late-onset diseases (Spearman's ρ = 0.32, p < 10-16); and (3) the disease onset age and heritability are negatively correlated (ρ = -0.46, p < 10-16).

Conjunction of factors triggering waves of seasonal influenza.

Using several longitudinal datasets describing putative factors affecting influenza incidence and clinical data on the disease and health status of over 150 million human subjects observed over a decade, we investigated the source and the mechanistic triggers of influenza epidemics. We conclude that the initiation of a pan-continental influenza wave emerges from the simultaneous realization of a complex set of conditions. The strongest predictor groups are as follows, ranked by importance: (1) the host population's socio- and ethno-demographic properties; (2) weather variables pertaining to specific humidity, temperature, and solar radiation; (3) the virus' antigenic drift over time; (4) the host population'€™s land-based travel habits, and; (5) recent spatio-temporal dynamics, as reflected in the influenza wave auto-correlation. The models we infer are demonstrably predictive (area under the Receiver Operating Characteristic curve 80%) when tested with out-of-sample data, opening the door to the potential formulation of new population-level intervention and mitigation policies.

Data smashing: uncovering lurking order in data.

From automatic speech recognition to discovering unusual stars, underlying almost all automated discovery tasks is the ability to compare and contrast data streams with each other, to identify connections and spot outliers. Despite the prevalence of data, however, automated methods are not keeping pace. A key bottleneck is that most data comparison algorithms today rely on a human expert to specify what 'features' of the data are relevant for comparison. Here, we propose a new principle for estimating the similarity between the sources of arbitrary data streams, using neither domain knowledge nor learning. We demonstrate the application of this principle to the analysis of data from a number of real-world challenging problems, including the disambiguation of electro-encephalograph patterns pertaining to epileptic seizures, detection of anomalous cardiac activity from heart sound recordings and classification of astronomical objects from raw photometry. In all these cases and without access to any domain knowledge, we demonstrate performance on a par with the accuracy achieved by specialized algorithms and heuristics devised by domain experts. We suggest that data smashing principles may open the door to understanding increasingly complex observations, especially when experts do not know what to look for.

Abductive learning of quantized stochastic processes with probabilistic finite automata.

We present an unsupervised learning algorithm (GenESeSS) to infer the causal structure of quantized stochastic processes, defined as stochastic dynamical systems evolving over discrete time, and producing quantized observations. Assuming ergodicity and stationarity, GenESeSS infers probabilistic finite state automata models from a sufficiently long observed trace. Our approach is abductive; attempting to infer a simple hypothesis, consistent with observations and modelling framework that essentially fixes the hypothesis class. The probabilistic automata we infer have no initial and terminal states, have no structural restrictions and are shown to be probably approximately correct-learnable. Additionally, we establish rigorous performance guarantees and data requirements, and show that GenESeSS correctly infers long-range dependencies. Modelling and prediction examples on simulated and real data establish relevance to automated inference of causal stochastic structures underlying complex physical phenomena.

Supervised self-organization of homogeneous swarms using ergodic projections of Markov chains.

This paper formulates a self-organization algorithm to address the problem of global behavior supervision in engineered swarms of arbitrarily large population sizes. The swarms considered in this paper are assumed to be homogeneous collections of independent identical finite-state agents, each of which is modeled by an irreducible finite Markov chain. The proposed algorithm computes the necessary perturbations in the local agents' behavior, which guarantees convergence to the desired observed state of the swarm. The ergodicity property of the swarm, which is induced as a result of the irreducibility of the agent models, implies that while the local behavior of the agents converges to the desired behavior only in the time average, the overall swarm behavior converges to the specification and stays there at all times. A simulation example illustrates the underlying concept.