Framework for converting mechanistic network models to probabilistic models

There are two prominent paradigms for the modelling of networks: in the first, referred to as the mechanistic approach, one specifies a set of domain-specific mechanistic rules that are used to grow or evolve the network over time; in the second, referred to as the probabilistic approach, one describes a model that specifies the likelihood of observing a given network. Mechanistic models (models developed based on the mechanistic approach) are appealing because they capture scientific processes that are believed to be responsible for network generation; however, they do not easily lend themselves to the use of inferential techniques when compared with probabilistic models. We introduce a general framework for converting a mechanistic network model (MNM) to a probabilistic network model (PNM). The proposed framework makes it possible to identify the essential network properties and their joint probability distribution for some MNMs; doing so makes it possible to address questions such as whether two different mechanistic models generate networks with identical distributions of properties, or whether a network property, such as clustering, is over- or under-represented in the networks generated by the model of interest compared with a reference model. The proposed framework is intended to bridge some of the gap that currently exists between the formulation and representation of mechanistic and PNMs. We also highlight limitations of PNMs that need to be addressed in order to close this gap.

Probabilistic Learning and Psychological Similarity

The notions of psychological similarity and probabilistic learning are key posits in cognitive, computational, and developmental psychology and in machine learning. However, their explanatory relationship is rarely made explicit within and across these research fields. This opinionated review critically evaluates how these notions can mutually inform each other within computational cognitive science. Using probabilistic models of concept learning as a case study, I argue that two notions of psychological similarity offer important normative constraints to guide modelers' interpretations of representational primitives. In particular, the two notions furnish probabilistic models of cognition with meaningful interpretations of what the associated subjective probabilities in the model represent and how they attach to experiences from which the agent learns. Similarity representations thereby provide probabilistic models with cognitive, as opposed to purely mathematical, content.

A Quantum Model of Trust Calibration in Human-AI Interactions

This exploratory study investigates a human agent's evolving judgements of reliability when interacting with an AI system. Two aims drove this investigation: (1) compare the predictive performance of quantum vs. Markov random walk models regarding human reliability judgements of an AI system and (2) identify a neural correlate of the perturbation of a human agent's judgement of the AI's reliability. As AI becomes more prevalent, it is important to understand how humans trust these technologies and how trust evolves when interacting with them. A mixed-methods experiment was developed for exploring reliability calibration in human-AI interactions. The behavioural data collected were used as a baseline to assess the predictive performance of the quantum and Markov models. We found the quantum model to better predict the evolving reliability ratings than the Markov model. This may be due to the quantum model being more amenable to represent the sometimes pronounced within-subject variability of reliability ratings. Additionally, a clear event-related potential response was found in the electroencephalographic (EEG) data, which is attributed to the expectations of reliability being perturbed. The identification of a trust-related EEG-based measure opens the door to explore how it could be used to adapt the parameters of the quantum model in real time.

Dynamic Compensation of a Piezoelectric Accelerometer Obtained through a General Probabilistic Approach

Dynamic compensation is the (partial) correction of the measurement signals for the effects due to bandwidth limitations of measurement systems and constitutes a research topic in dynamic measurement. The dynamic compensation of an accelerometer is here considered, as obtained by a method that directly comes from a general probabilistic model of the measurement process. Although the application of the method is simple, the analytical development of the corresponding compensation filter is quite complex and had been previously developed only for first-order systems, whilst here a second-order system is considered, thus moving from a scalar to a vector problem. The effectiveness of the method has been tested both through simulation and by a dedicated experiment. Both tests have shown the capability of the method of significantly improve the performance of the measurement system when dynamic effects are more prevalent than additive observation noise.

Dynamic Bayesian Model for Detecting Obstructive Respiratory Events by Using an Experimental Model

In this study, we propose a model-based tool for the detection of obstructive apnea episodes by using ECG features from a single lead channel. Several sequences of recurrent apnea were provoked in separate 15-min periods in anesthetized rats during an experimental model of obstructive sleep apnea (OSA). Morphology-based ECG markers and the beat-to-beat interval (RR) were assessed in each sequence. These markers were used to train dynamic Bayesian networks (DBN) with different orders and feature combinations to find a good tradeoff between network complexity and apnea-detection performance. By using a filtering approach, the resulting DBNs were used to infer the apnea probability signal for subsequent episodes in the same rat. These signals were then processed using by 15-s epochs to determine whether epochs were classified as apneic or nonapneic. Our results showed that fifth-order models provided suitable RMSE values, since higher order models become significantly more complex and present worse generalization. A global threshold of 0.2 gave the best overall performance for all combinations tested, with Acc = 81.3%, Se = 69.8% and Sp = 81.5%, using only two parameters including the RR and Ds (R-wave downslope) markers. We concluded that multivariate models using DBNs represent a powerful tool for detecting obstructive apnea episodes in short segments, which may also serve to estimate the number of total events in a given time period.

Exploiting Domain Knowledge as Causal Independencies in Modeling Gestational Diabetes

We consider the problem of modeling gestational diabetes in a clinical study and develop a domain expert-guided probabilistic model that is both interpretable and explainable. Specifically, we construct a probabilistic model based on causal independence (Noisy-Or) from a carefully chosen set of features. We validate the efficacy of the model on the clinical study and demonstrate the importance of the features and the causal independence model.

Prospective cohort study of psoriatic arthritis risk in patients with psoriasis in a real-world psoriasis registry

BACKGROUND

The characteristics that predict the onset of psoriatic arthritis (PsA) among patients with psoriasis (PsO) may inform diagnosis and treatment.

OBJECTIVE

To develop a model to predict the 2-year risk of developing PsA among patients with PsO.

METHODS

This was a prospective cohort study of patients in the CorEvitas Psoriasis Registry without PsA at enrollment and with 24-month follow-up. Unregularized and regularized logistic regression models were developed and tested using descriptive variables to predict dermatologist-identified PsA at 24 months. Model performance was compared using the area under the receiver operating characteristic curve (AUC), sensitivity, and specificity.

RESULTS

A total of 1489 patients were included. Nine unique predictive models were developed and tested. The optimal model, including Psoriasis Epidemiology Screening Tool (PEST), body mass index (BMI), modified Rheumatic Disease Comorbidity Index, work status, alcohol use, and patient-reported fatigue, predicted the onset of PsA within 24 months (AUC = 68.9%, sensitivity = 82.9%, specificity = 48.8%). A parsimonious model including PEST and BMI had similar performance (AUC = 68.8%; sensitivity = 92.7%, specificity = 36.5%).

LIMITATIONS

PsA misclassification bias by dermatologists.

CONCLUSION

PEST and BMI were important factors in predicting the development of PsA in patients with PsO over 2 years and thereby foundational for future PsA risk model development.

Beyond the Backbone: The Next Generation of Pathwalking Utilities for Model Building in CryoEM Density Maps

Single-particle electron cryomicroscopy (cryoEM) has become an indispensable tool for studying structure and function in macromolecular assemblies. As an integral part of the cryoEM structure determination process, computational tools have been developed to build atomic models directly from a density map without structural templates. Nearly a decade ago, we created Pathwalking, a tool for de novo modeling of protein structure in near-atomic resolution cryoEM density maps. Here, we present the latest developments in Pathwalking, including the addition of probabilistic models, as well as a companion tool for modeling waters and ligands. This software was evaluated on the 2021 CryoEM Ligand Challenge density maps, in addition to identifying ligands in three IP3R1 density maps at ~3 Å to 4.1 Å resolution. The results clearly demonstrate that the Pathwalking de novo modeling pipeline can construct accurate protein structures and reliably localize and identify ligand density directly from a near-atomic resolution map.

Mapping the evolution of T cell states during response and resistance to adoptive cellular therapy

To elucidate mechanisms by which T cells eliminate leukemia, we study donor lymphocyte infusion (DLI), an established immunotherapy for relapsed leukemia. We model T cell dynamics by integrating longitudinal, multimodal data from 94,517 bone marrow-derived single T cell transcriptomes in addition to chromatin accessibility and single T cell receptor sequencing from patients undergoing DLI. We find that responsive tumors are defined by enrichment of late-differentiated T cells before DLI and rapid, durable expansion of early differentiated T cells after treatment, highly similar to "terminal" and "precursor" exhausted subsets, respectively. Resistance, in contrast, is defined by heterogeneous T cell dysfunction. Surprisingly, early differentiated T cells in responders mainly originate from pre-existing and novel clonotypes recruited to the leukemic microenvironment, rather than the infusion. Our work provides a paradigm for analyzing longitudinal single-cell profiling of scenarios beyond adoptive cell therapy and introduces Symphony, a Bayesian approach to infer regulatory circuitry underlying T cell subsets, with broad relevance to exhaustion antagonists across cancers.

Directional Gaussian Mixture Models of the Gut Microbiome Elucidate Microbial Spatial Structure

The gut microbiome is spatially heterogeneous, with environmental niches contributing to the distribution and composition of microbial populations. A recently developed mapping technology, MaPS-seq, aims to characterize the spatial organization of the gut microbiome by providing data about local microbial populations. However, information about the global arrangement of these populations is lost by MaPS-seq. To address this, we propose a class of Gaussian mixture models (GMM) with spatial dependencies between mixture components in order to computationally recover the relative spatial arrangement of microbial communities. We demonstrate on synthetic data that our spatial models can identify global spatial dynamics, accurately cluster data, and improve parameter inference over a naive GMM. We applied our model to three MaPS-seq data sets taken from various regions of the mouse intestine. On cecal and distal colon data sets, we find our model accurately recapitulates known spatial behaviors of the gut microbiome, including compositional differences between mucus and lumen-associated populations. Our model also seems to capture the role of a pH gradient on microbial populations in the mouse ileum and proposes new behaviors as well. IMPORTANCE The spatial arrangement of the microbes in the gut microbiome is a defining characteristic of its behavior. Various experimental studies have attempted to provide glimpses into the mechanisms that contribute to microbial arrangements. However, many of these descriptions are qualitative. We developed a computational method that takes microbial spatial data and learns many of the experimentally validated spatial factors. We can then use our model to propose previously unknown spatial behaviors. Our results demonstrate that the gut microbiome, while exceptionally large, has predictable spatial patterns that can be used to help us understand its role in health and disease.

The Role of Instrumental Variables in Causal Inference Based on Independence of Cause and Mechanism

Causal inference methods based on conditional independence construct Markov equivalent graphs and cannot be applied to bivariate cases. The approaches based on independence of cause and mechanism state, on the contrary, that causal discovery can be inferred for two observations. In our contribution, we pose a challenge to reconcile these two research directions. We study the role of latent variables such as latent instrumental variables and hidden common causes in the causal graphical structures. We show that methods based on the independence of cause and mechanism indirectly contain traces of the existence of the hidden instrumental variables. We derive a novel algorithm to infer causal relationships between two variables, and we validate the proposed method on simulated data and on a benchmark of cause-effect pairs. We illustrate by our experiments that the proposed approach is simple and extremely competitive in terms of empirical accuracy compared to the state-of-the-art methods.

A predictive model for fracture in human ribs based on in vitro acoustic emission data

PURPOSE

The aim of this paper is to propose a fracture model for human ribs based on acoustic emission (AE) data. The accumulation of microcracking until a macroscopic crack is produced can be monitored by AE. The macrocrack propagation causes the loss of the structural integrity of the rib.

METHODS

The AE technique was used in in vitro bending tests of human ribs. The AE data obtained were used to construct a quantitative model that allows an estimation of the failure stress from the signals detected. The model predicts the ultimate stress with an error of less than 3.5% (even at stresses 15% lower than failure stress), which makes it possible to safely anticipate the failure of the rib.

RESULTS

The percolation theory was used to model crack propagation. Moreover, a quantitative probability-based model for the expected number of AE signals has been constructed, incorporating some ideas of percolation theory. The model predicts that AE signals associated with micro-failures should exhibit a vertical asymptote when stress increases. The occurrence of this vertical asymptote was attested in our experimental observations. The total number of microfailures detected prior to the failure is N ≈ 100 and the ultimate stress is σ ∞ = 197 ± 62 MPa. A significant correlation (p < 0.0001) between σ ∞ and the predicted value is found, using only the first N = 30 micro-failures (correlation improves for N higher).

CONCLUSIONS

The measurements and the shape of the curves predicted by the model fit well. In addition, the model parameters seem to explain quantitatively and qualitatively the distribution of the AE signals as the material approaches the macroscopic fracture. Moreover, some of these parameters correlate with anthropometric variables, such as age or Body Mass Index. The proposed model could be used to predict the structural failure of ribs subjected to bending.

Modeling the heterogeneity in COVID-19's reproductive number and its impact on predictive scenarios

The correct evaluation of the reproductive number R for COVID-19 is central in the quantification of the potential scope of the pandemic and the selection of an appropriate course of action. In most models, R is modeled as a constant - effectively averaging out the inherent variability of the transmission process due to varying individual contact rates, population densities, or temporal factors amongst many. Yet, due to the exponential nature of epidemic growth, the error due to this simplification can be rapidly amplified, and its extent remains unknown. How can this intrinsic variability be percolated into epidemic models, and its impact, better quantified? We study this question here through a Bayesian perspective that captures at scale the heterogeneity of a population and environmental conditions, creating a bridge between the traditional agent-based and compartmental approaches. We use our model to simulate the spread as well as the impact of different social distancing strategies on real COVID-19 data, and highlight the significant impact of the heterogeneity. We emphasize that the contribution of this paper focuses on discussing the importance of the impact of R's heterogeneity on uncertainty quantification from a statistical viewpoint, rather than developing new predictive models.

Predicting Emotional States Using Behavioral Markers Derived From Passively Sensed Data: Data-Driven Machine Learning Approach

BACKGROUND

Mental health disorders affect multiple aspects of patients' lives, including mood, cognition, and behavior. eHealth and mobile health (mHealth) technologies enable rich sets of information to be collected noninvasively, representing a promising opportunity to construct behavioral markers of mental health. Combining such data with self-reported information about psychological symptoms may provide a more comprehensive and contextualized view of a patient's mental state than questionnaire data alone. However, mobile sensed data are usually noisy and incomplete, with significant amounts of missing observations. Therefore, recognizing the clinical potential of mHealth tools depends critically on developing methods to cope with such data issues.

OBJECTIVE

This study aims to present a machine learning-based approach for emotional state prediction that uses passively collected data from mobile phones and wearable devices and self-reported emotions. The proposed methods must cope with high-dimensional and heterogeneous time-series data with a large percentage of missing observations.

METHODS

Passively sensed behavior and self-reported emotional state data from a cohort of 943 individuals (outpatients recruited from community clinics) were available for analysis. All patients had at least 30 days' worth of naturally occurring behavior observations, including information about physical activity, geolocation, sleep, and smartphone app use. These regularly sampled but frequently missing and heterogeneous time series were analyzed with the following probabilistic latent variable models for data averaging and feature extraction: mixture model (MM) and hidden Markov model (HMM). The extracted features were then combined with a classifier to predict emotional state. A variety of classical machine learning methods and recurrent neural networks were compared. Finally, a personalized Bayesian model was proposed to improve performance by considering the individual differences in the data and applying a different classifier bias term for each patient.

RESULTS

Probabilistic generative models proved to be good preprocessing and feature extractor tools for data with large percentages of missing observations. Models that took into account the posterior probabilities of the MM and HMM latent states outperformed those that did not by more than 20%, suggesting that the underlying behavioral patterns identified were meaningful for individuals' overall emotional state. The best performing generalized models achieved a 0.81 area under the curve of the receiver operating characteristic and 0.71 area under the precision-recall curve when predicting self-reported emotional valence from behavior in held-out test data. Moreover, the proposed personalized models demonstrated that accounting for individual differences through a simple hierarchical model can substantially improve emotional state prediction performance without relying on previous days' data.

CONCLUSIONS

These findings demonstrate the feasibility of designing machine learning models for predicting emotional states from mobile sensing data capable of dealing with heterogeneous data with large numbers of missing observations. Such models may represent valuable tools for clinicians to monitor patients' mood states.

Bayesian network to predict hepatitis B surface antigen seroclearance in chronic hepatitis B patients

There is a need for an interpretable, accurate and interactions-considered model for predicting hepatitis B surface antigen (HBsAg) seroclearance. We aimed to construct a Bayesian network (BN) model using available medical records to predict HBsAg seroclearance in chronic hepatitis B (CHB) patients, and to evaluate the model's performance. This was a case-control study. A total of 1966 consecutive CHB patients (mean age 39.04 ± 11.23 years) between January 2006 and June 2015 were included. The demographic and clinical characteristics, laboratory data and imaging parameters were obtained and used to build a BN model to estimate the probability of HBsAg seroclearance. Baseline serum HBsAg and hepatitis Be antigen (HBeAg) levels, virological response and HBeAg seroclearance were the most significant predictors of HBsAg seroclearance. The post-test probability table showed that patients with baseline HBsAg concentrations ≤2000 IU/mL, negative baseline HBeAg, an initial virological response and without HBeAg seroclearance (i.e. no recurrence of HBeAg positivity during follow-up) were most likely to have HBsAg seroclearance. The constructed BN model had an area under the receiver operating characteristic curves of 0.896 (95% confidence interval [CI]: 0.892, 0.899), a sensitivity of 0.840 (95% CI: 0.833, 0.846), a specificity of 0.880 (95% CI: 0.876, 0.884) and an accuracy of 0.878 (95% CI: 0.874, 0.882) for predicting HBsAg seroclearance. The established BN model accurately estimated the probability of HBsAg seroclearance and is a promising tool to assist clinical decision-making.

Dissecting heterogeneous cell populations across drug and disease conditions with PopAlign

Single-cell measurement techniques can now probe gene expression in heterogeneous cell populations from the human body across a range of environmental and physiological conditions. However, new mathematical and computational methods are required to represent and analyze gene-expression changes that occur in complex mixtures of single cells as they respond to signals, drugs, or disease states. Here, we introduce a mathematical modeling platform, PopAlign, that automatically identifies subpopulations of cells within a heterogeneous mixture and tracks gene-expression and cell-abundance changes across subpopulations by constructing and comparing probabilistic models. Probabilistic models provide a low-error, compressed representation of single-cell data that enables efficient large-scale computations. We apply PopAlign to analyze the impact of 40 different immunomodulatory compounds on a heterogeneous population of donor-derived human immune cells as well as patient-specific disease signatures in multiple myeloma. PopAlign scales to comparisons involving tens to hundreds of samples, enabling large-scale studies of natural and engineered cell populations as they respond to drugs, signals, or physiological change.

Prediction of the Transition From Subexponential to the Exponential Transmission of SARS-CoV-2 in Chennai, India: Epidemic Nowcasting

Background

Several countries adopted lockdown to slowdown the exponential transmission of the coronavirus disease (COVID-19) epidemic. Disease transmission models and the epidemic forecasts at the national level steer the policy to implement appropriate intervention strategies and budgeting. However, it is critical to design a data-driven reliable model for nowcasting for smaller populations, in particular metro cities.

Objective

The aim of this study is to analyze the transition of the epidemic from subexponential to exponential transmission in the Chennai metro zone and to analyze the probability of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) secondary infections while availing the public transport systems in the city.

Methods

A single geographical zone “Chennai-Metro-Merge” was constructed by combining Chennai District with three bordering districts. Subexponential and exponential models were developed to analyze and predict the progression of the COVID-19 epidemic. Probabilistic models were applied to assess the probability of secondary infections while availing public transport after the release of the lockdown.

Results

The model predicted that transition from subexponential to exponential transmission occurs around the eighth week after the reporting of a cluster of cases. The probability of secondary infections with a single index case in an enclosure of the city bus, the suburban train general coach, and the ladies coach was found to be 0.192, 0.074, and 0.114, respectively.

Conclusions

Nowcasting at the early stage of the epidemic predicts the probable time point of the exponential transmission and alerts the public health system. After the lockdown release, public transportation will be the major source of SARS-CoV-2 transmission in metro cities, and appropriate strategies based on nowcasting are needed.

Quantitative Predictive Modelling Approaches to Understanding Rheumatoid Arthritis: A Brief Review

Rheumatoid arthritis is a chronic autoimmune disease that is a major public health challenge. The disease is characterised by inflammation of synovial joints and cartilage erosion, which lead to chronic pain, poor life quality and, in some cases, mortality. Understanding the biological mechanisms behind the progression of the disease, as well as developing new methods for quantitative predictions of disease progression in the presence/absence of various therapies is important for the success of therapeutic approaches. The aim of this study is to review various quantitative predictive modelling approaches for understanding rheumatoid arthritis. To this end, we start by briefly discussing the biology of this disease and some current treatment approaches, as well as emphasising some of the open problems in the field. Then, we review various mathematical mechanistic models derived to address some of these open problems. We discuss models that investigate the biological mechanisms behind the progression of the disease, as well as pharmacokinetic and pharmacodynamic models for various drug therapies. Furthermore, we highlight models aimed at optimising the costs of the treatments while taking into consideration the evolution of the disease and potential complications.

Development of a Taekwondo Combat Model Based on Markov Analysis

The purpose of the present study was to examine male and female Olympic taekwondo competitors' movement patterns according to their tactical actions by applying a Markov processes analysis. To perform this study, 11,474 actions by male competitors and 12,980 actions by female competitors were compiled and analyzed. The results yielded 32 significant sequences among male competitors and 30 among female competitors. Male competitors demonstrated 11 sequences initiated by an attack, 11 initiated by a counterattack, and 10 initiated by a defensive action. Female competitors demonstrated nine sequences initiated by an attack, 11 initiated by a counterattack, and 10 initiated by a defensive move. The five most popular sequences were the opening and dodge, the direct attack and simultaneous counterattack, the dodge with a direct attack, the indirect attack and simultaneous counterattack, and the simultaneous counterattack with a direct attack. Markov chains help provide coaches and researchers with relevant information about the frequency of actions, both in terms of their frequency of occurrence and the order of their occurrence, during a real competition. It is suggested that coaches and athletes focus on these patterns when training for a real competition.

Visualizing probabilistic models and data with Intensive Principal Component Analysis

Unsupervised learning makes manifest the underlying structure of data without curated training and specific problem definitions. However, the inference of relationships between data points is frustrated by the "curse of dimensionality" in high dimensions. Inspired by replica theory from statistical mechanics, we consider replicas of the system to tune the dimensionality and take the limit as the number of replicas goes to zero. The result is intensive embedding, which not only is isometric (preserving local distances) but also allows global structure to be more transparently visualized. We develop the Intensive Principal Component Analysis (InPCA) and demonstrate clear improvements in visualizations of the Ising model of magnetic spins, a neural network, and the dark energy cold dark matter ([Formula: see text]) model as applied to the cosmic microwave background.

Detecting adaptive convergent amino acid evolution

In evolutionary genomics, researchers have taken an interest in identifying substitutions that subtend convergent phenotypic adaptations. This is a difficult question that requires distinguishing foreground convergent substitutions that are involved in the convergent phenotype from background convergent substitutions. Those may be linked to other adaptations, may be neutral or may be the consequence of mutational biases. Furthermore, there is no generally accepted definition of convergent substitutions. Various methods that use different definitions have been proposed in the literature, resulting in different sets of candidate foreground convergent substitutions. In this article, we first describe the processes that can generate foreground convergent substitutions in coding sequences, separating adaptive from non-adaptive processes. Second, we review methods that have been proposed to detect foreground convergent substitutions in coding sequences and expose the assumptions that underlie them. Finally, we examine their power on simulations of convergent changes—including in the presence of a change in the efficacy of selection—and on empirical alignments.

This article is part of the theme issue ‘Convergent evolution in the genomics era: new insights and directions'.

Exploring the mechanisms behind the country-specific time of Zika virusimportation

The international spread of Zika virus (ZIKV) began in Brazil in 2015. To estimate the riskof observing imported ZIKV cases, we calculated effective distance, typically an excellent predictorof arrival time, from airline network data. However, we eventually concluded that, for ZIKV, effectivedistance alone is not an adequate predictor of arrival time, which we partly attributed to the difficultyof diagnosing and ascertaining ZIKV infections. Herein, we explored the mechanisms behind theobserved time delay of ZIKV importation by country, statistically decomposing the delay into twoparts: the actual time to importation from Brazil and the reporting delay. The latter was modeled as afunction of the gross domestic product (GDP) and other variables that influence underlying diagnosticcapacity in a given country. We showed that a high GDP per capita is a good predictor of shortreporting delay. ZIKV infection is generally mild and, without substantial laboratory capacity, casescan be underestimated. This study successfully demonstrates this phenomenon and emphasizes theimportance of accounting for reporting delays as part of the data generating process for estimatingtime to importation.

Analysing Cooking Behaviour in Home Settings: Towards Health Monitoring

Wellbeing is often affected by health-related conditions. Among them are nutrition-related health conditions, which can significantly decrease the quality of life. We envision a system that monitors the kitchen activities of patients and that based on the detected eating behaviour could provide clinicians with indicators for improving a patient’s health. To be successful, such system has to reason about the person’s actions and goals. To address this problem, we introduce a symbolic behaviour recognition approach, called Computational Causal Behaviour Models (CCBM). CCBM combines symbolic representation of person’s behaviour with probabilistic inference to reason about one’s actions, the type of meal being prepared, and its potential health impact. To evaluate the approach, we use a cooking dataset of unscripted kitchen activities, which contains data from various sensors in a real kitchen. The results show that the approach is able to reason about the person’s cooking actions. It is also able to recognise the goal in terms of type of prepared meal and whether it is healthy. Furthermore, we compare CCBM to state-of-the-art approaches such as Hidden Markov Models (HMM) and decision trees (DT). The results show that our approach performs comparable to the HMM and DT when used for activity recognition. It outperformed the HMM for goal recognition of the type of meal with median accuracy of 1 compared to median accuracy of 0.12 when applying the HMM. Our approach also outperformed the HMM for recognising whether a meal is healthy with a median accuracy of 1 compared to median accuracy of 0.5 with the HMM.

Cost Effectiveness of Intracranial Pressure Monitoring in Pediatric Patients with Severe Traumatic Brain Injury: A Simulation Modeling Approach

OBJECTIVES

To conduct an economic evaluation of intracranial pressure (ICP) monitoring on the basis of current evidence from pediatric patients with severe traumatic brain injury, through a statistical model.

METHODS

The statistical model is a decision tree, whose branches take into account the severity of the lesion, the hospitalization costs, and the quality-adjusted life-year for the first 6 months post-trauma. The inputs consist of probability distributions calculated from a sample of 33 surviving children with severe traumatic brain injury, divided into two groups: with ICP monitoring (monitoring group) and without ICP monitoring (control group). The uncertainty of the parameters from the sample was quantified through a probabilistic sensitivity analysis using the Monte-Carlo simulation method. The model overcomes the drawbacks of small sample sizes, unequal groups, and the ethical difficulty in randomly assigning patients to a control group (without monitoring).

RESULTS

The incremental cost in the monitoring group was Mex$3,934 (Mexican pesos), with an increase in quality-adjusted life-year of 0.05. The incremental cost-effectiveness ratio was Mex$81,062. The cost-effectiveness acceptability curve had a maximum at 54% of the cost effective iterations. The incremental net health benefit for a willingness to pay equal to 1 time the per capita gross domestic product for Mexico was 0.03, and the incremental net monetary benefit was Mex$5,358.

CONCLUSIONS

The results of the model suggest that ICP monitoring is cost effective because there was a monetary gain in terms of the incremental net monetary benefit.

Reducing Energy Consumption in Serial Production Lines with Bernoulli Reliability Machines

In this paper, an integrated model to minimize energy consumption while maintaining desired productivity in Bernoulli serial lines is introduced. Exact analysis of optimal allocation of production capacity is carried out for small systems, such as three- and four-machine lines with small buffers. For medium size systems (e.g., three- and four-machine lines with larger buffers, or five-machine lines with small buffers), an aggregation procedure is introduced to evaluate line production rate, and then use it to search optimal allocation of machine efficiency to minimize energy usage. Insights and allocation principles are obtained through the analyses. Finally, for larger systems, a heuristic algorithm is proposed and validated through extensive numerical experiments.

The Evolution of Haploid Chromosome Numbers in the Sunflower Family

Chromosome number changes during the evolution of angiosperms are likely to have played a major role in speciation. Their study is of utmost importance, especially now, as a probabilistic model is available to study chromosome evolution within a phylogenetic framework. In the present study, likelihood models of chromosome number evolution were fitted to the largest family of flowering plants, the Asteraceae. Specifically, a phylogenetic supertree of this family was used to reconstruct the ancestral chromosome number and infer genomic events. Our approach inferred that the ancestral chromosome number of the family is n = 9. Also, according to the model that best explained our data, the evolution of haploid chromosome numbers in Asteraceae was a very dynamic process, with genome duplications and descending dysploidy being the most frequent genomic events in the evolution of this family. This model inferred more than one hundred whole genome duplication events; however, it did not find evidence for a paleopolyploidization at the base of this family, which has previously been hypothesized on the basis of sequence data from a limited number of species. The obtained results and potential causes of these discrepancies are discussed.

Data-directed RNA secondary structure prediction using probabilistic modeling

Structure dictates the function of many RNAs, but secondary RNA structure analysis is either labor intensive and costly or relies on computational predictions that are often inaccurate. These limitations are alleviated by integration of structure probing data into prediction algorithms. However, existing algorithms are optimized for a specific type of probing data. Recently, new chemistries combined with advances in sequencing have facilitated structure probing at unprecedented scale and sensitivity. These novel technologies and anticipated wealth of data highlight a need for algorithms that readily accommodate more complex and diverse input sources. We implemented and investigated a recently outlined probabilistic framework for RNA secondary structure prediction and extended it to accommodate further refinement of structural information. This framework utilizes direct likelihood-based calculations of pseudo-energy terms per considered structural context and can readily accommodate diverse data types and complex data dependencies. We use real data in conjunction with simulations to evaluate performances of several implementations and to show that proper integration of structural contexts can lead to improvements. Our tests also reveal discrepancies between real data and simulations, which we show can be alleviated by refined modeling. We then propose statistical preprocessing approaches to standardize data interpretation and integration into such a generic framework. We further systematically quantify the information content of data subsets, demonstrating that high reactivities are major drivers of SHAPE-directed predictions and that better understanding of less informative reactivities is key to further improvements. Finally, we provide evidence for the adaptive capability of our framework using mock probe simulations.

Estimating the value of medical treatments to patients using probabilistic multi criteria decision analysis

BACKGROUND

Estimating the value of medical treatments to patients is an essential part of healthcare decision making, but is mostly done implicitly and without consulting patients. Multi criteria decision analysis (MCDA) has been proposed for the valuation task, while stated preference studies are increasingly used to measure patient preferences. In this study we propose a methodology for using stated preferences to weigh clinical evidence in an MCDA model that includes uncertainty in both patient preferences and clinical evidence explicitly.

METHODS

A probabilistic MCDA model with an additive value function was developed and illustrated using a case on hypothetical treatments for depression. The patient-weighted values were approximated with Monte Carlo simulations and compared to expert-weighted results. Decision uncertainty was calculated as the probability of rank reversal for the first rank. Furthermore, scenario analyses were done to assess the relative impact of uncertainty in preferences and clinical evidence, and of assuming uniform preference distributions.

RESULTS

The patient-weighted values for drug A, drug B, drug C, and placebo were 0.51 (95% CI: 0.48 to 0.54), 0.51 (95% CI: 0.48 to 0.54), 0.54 (0.49 to 0.58), and 0.15 (95% CI: 0.13 to 0.17), respectively. Drug C was the most preferred treatment and the rank reversal probability for first rank was 27%. This probability decreased to 18% when uncertainty in performances was not included and increased to 41% when uncertainty in criterion weights was not included. With uniform preference distributions, the first rank reversal probability increased to 61%. The expert-weighted values for drug A, drug B, drug C, and placebo were 0.67 (95% CI: 0.65 to 0.68), 0.57 (95% CI: 0.56 to 0.59), 0.67 (95% CI: 0.61 to 0.71), and 0.19 (95% CI: 0.17 to 0.21). The rank reversal probability for the first rank according to experts was 49%.

CONCLUSIONS

Preferences elicited from patients can be used to weigh clinical evidence in a probabilistic MCDA model. The resulting treatment values can be contrasted to results from experts, and the impact of uncertainty can be quantified using rank probabilities. Future research should focus on integrating the model with regulatory decision frameworks and on including other types of uncertainty.

Thresholds of species loss in Amazonian deforestation frontier landscapes

In the Brazilian Amazon, private land accounts for the majority of remaining native vegetation. Understanding how land-use change affects the composition and distribution of biodiversity in farmlands is critical for improving conservation strategies in the face of rapid agricultural expansion. Working across an area exceeding 3 million ha in the southwestern state of Rondônia, we assessed how the extent and configuration of remnant forest in replicate 10,000-ha landscapes has affected the occurrence of a suite of Amazonian mammals and birds. In each of 31 landscapes, we used field sampling and semistructured interviews with landowners to determine the presence of 28 large and medium sized mammals and birds, as well as a further 7 understory birds. We then combined results of field surveys and interviews with a probabilistic model of deforestation. We found strong evidence for a threshold response of sampled biodiversity to landscape level forest cover; landscapes with <30-40% forest cover hosted markedly fewer species. Results from field surveys and interviews yielded similar thresholds. These results imply that in partially deforested landscapes many species are susceptible to extirpation following relatively small additional reductions in forest area. In the model of deforestation by 2030 the number of 10,000-ha landscapes under a conservative threshold of 43% forest cover almost doubled, such that only 22% of landscapes would likely to be able to sustain at least 75% of the 35 focal species we sampled. Brazilian law requires rural property owners in the Amazon to retain 80% forest cover, although this is rarely achieved. Prioritizing efforts to ensure that entire landscapes, rather than individual farms, retain at least 50% forest cover may help safeguard native biodiversity in private forest reserves in the Amazon.

The development of the ProMAS: a Probabilistic Medication Adherence Scale

Current self-report medication adherence measures often provide heavily skewed results with limited variance, suggesting that most participants are highly adherent. This contrasts with findings from objective adherence measures. We argue that one of the main limitations of these self-report measures is the limited range covered by the behaviors assessed. That is, the items do not match the adherence behaviors that people perform, resulting in a ceiling effect. In this paper, we present a new self-reported medication adherence scale based on the Rasch model approach (the ProMAS), which covers a wide range of adherence behaviors. The ProMAS was tested with 370 elderly receiving medication for chronic conditions. The results indicated that the ProMAS provided adherence scores with sufficient fit to the Rasch model. Furthermore, the ProMAS covered a wider range of adherence behaviors compared to the widely used Medication Adherence Report Scale (MARS) instrument, resulting in more variance and less skewness in adherence scores. We conclude that the ProMAS is more capable of discriminating between people with different adherence rates than the MARS.

Does time ever fly or slow down? The difficult interpretation of psychophysical data on time perception

Time perception is studied with subjective or semi-objective psychophysical methods. With subjective methods, observers provide quantitative estimates of duration and data depict the psychophysical function relating subjective duration to objective duration. With semi-objective methods, observers provide categorical or comparative judgments of duration and data depict the psychometric function relating the probability of a certain judgment to objective duration. Both approaches are used to study whether subjective and objective time run at the same pace or whether time flies or slows down under certain conditions. We analyze theoretical aspects affecting the interpretation of data gathered with the most widely used semi-objective methods, including single-presentation and paired-comparison methods. For this purpose, a formal model of psychophysical performance is used in which subjective duration is represented via a psychophysical function and the scalar property. This provides the timing component of the model, which is invariant across methods. A decisional component that varies across methods reflects how observers use subjective durations to make judgments and give the responses requested under each method. Application of the model shows that psychometric functions in single-presentation methods are uninterpretable because the various influences on observed performance are inextricably confounded in the data. In contrast, data gathered with paired-comparison methods permit separating out those influences. Prevalent approaches to fitting psychometric functions to data are also discussed and shown to be inconsistent with widely accepted principles of time perception, implicitly assuming instead that subjective time equals objective time and that observed differences across conditions do not reflect differences in perceived duration but criterion shifts. These analyses prompt evidence-based recommendations for best methodological practice in studies on time perception.

Modelling the efficiency of local versus central provision of intravenous thrombolysis after acute ischemic stroke

BACKGROUND AND PURPOSE

Prehospital redirection of stroke patients to a regional center is used as a strategy to maximize the provision of intravenous thrombolysis. We developed a model to quantify the benefit of redirection away from local services that were already providing thrombolysis.

METHODS

A microsimulation using hospital and ambulance data from consecutive emergency admissions to 10 local acute stroke units estimated the effect of redirection to 2 regional neuroscience centers. Modeled outcomes reflected additional journey time and accuracy of stroke identification in the prehospital phase, and the relative efficiency of patient selection and door-needle time for each local site compared with the nearest regional neuroscience center.

RESULTS

Thrombolysis was received by 223/1884 emergency admissions. Based on observed site performance, 68 additional patients would have been treated after theoretical redirection of 1269 true positive cases and 363 stroke mimics to the neuroscience center. Over 5 years redirection of this cohort generated 12.6 quality-adjusted life years at a marginal cost of £6730 ($10,320, €8347). The average additional cost of a quality-adjusted life year gain was £534 ($819, €673).

CONCLUSIONS

Under these specific circumstances, redirection would have improved outcomes from thrombolysis at little additional cost.

The four ingredients of single-sequence RNA secondary structure prediction. A unifying perspective

Any method for RNA secondary structure prediction is determined by four ingredients. The architecture is the choice of features implemented by the model (such as stacked basepairs, loop length distributions, etc.). The architecture determines the number of parameters in the model. The scoring scheme is the nature of those parameters (whether thermodynamic, probabilistic, or weights). The parameterization stands for the specific values assigned to the parameters. These three ingredients are referred to as "the model." The fourth ingredient is the folding algorithms used to predict plausible secondary structures given the model and the sequence of a structural RNA. Here, I make several unifying observations drawn from looking at more than 40 years of methods for RNA secondary structure prediction in the light of this classification. As a final observation, there seems to be a performance ceiling that affects all methods with complex architectures, a ceiling that impacts all scoring schemes with remarkable similarity. This suggests that modeling RNA secondary structure by using intrinsic sequence-based plausible "foldability" will require the incorporation of other forms of information in order to constrain the folding space and to improve prediction accuracy. This could give an advantage to probabilistic scoring systems since a probabilistic framework is a natural platform to incorporate different sources of information into one single inference problem.

A novel human autonomy assessment system

This article presents a novel human autonomy assessment system for generating context and discovering the behaviors of older people who use ambulant services. Our goal is to assist caregivers in assessing possibly abnormal health conditions in their clients concerning their level of autonomy, thus enabling caregivers to take countermeasures as soon as possible.

A range of complex probabilistic models for RNA secondary structure prediction that includes the nearest-neighbor model and more

The standard approach for single-sequence RNA secondary structure prediction uses a nearest-neighbor thermodynamic model with several thousand experimentally determined energy parameters. An attractive alternative is to use statistical approaches with parameters estimated from growing databases of structural RNAs. Good results have been reported for discriminative statistical methods using complex nearest-neighbor models, including CONTRAfold, Simfold, and ContextFold. Little work has been reported on generative probabilistic models (stochastic context-free grammars [SCFGs]) of comparable complexity, although probabilistic models are generally easier to train and to use. To explore a range of probabilistic models of increasing complexity, and to directly compare probabilistic, thermodynamic, and discriminative approaches, we created TORNADO, a computational tool that can parse a wide spectrum of RNA grammar architectures (including the standard nearest-neighbor model and more) using a generalized super-grammar that can be parameterized with probabilities, energies, or arbitrary scores. By using TORNADO, we find that probabilistic nearest-neighbor models perform comparably to (but not significantly better than) discriminative methods. We find that complex statistical models are prone to overfitting RNA structure and that evaluations should use structurally nonhomologous training and test data sets. Overfitting has affected at least one published method (ContextFold). The most important barrier to improving statistical approaches for RNA secondary structure prediction is the lack of diversity of well-curated single-sequence RNA secondary structures in current RNA databases.

Hybrid Modeling of Cell Signaling and Transcriptional Reprogramming and Its Application in C. elegans Development

Modeling of signal driven transcriptional reprogramming is critical for understanding of organism development, human disease, and cell biology. Many current modeling techniques discount key features of the biological sub-systems when modeling multiscale, organism-level processes. We present a mechanistic hybrid model, GESSA, which integrates a novel pooled probabilistic Boolean network model of cell signaling and a stochastic simulation of transcription and translation responding to a diffusion model of extracellular signals. We apply the model to simulate the well studied cell fate decision process of the vulval precursor cells (VPCs) in C. elegans, using experimentally derived rate constants wherever possible and shared parameters to avoid overfitting. We demonstrate that GESSA recovers (1) the effects of varying scaffold protein concentration on signal strength, (2) amplification of signals in expression, (3) the relative external ligand concentration in a known geometry, and (4) feedback in biochemical networks. We demonstrate that setting model parameters based on wild-type and LIN-12 loss-of-function mutants in C. elegans leads to correct prediction of a wide variety of mutants including partial penetrance of phenotypes. Moreover, the model is relatively insensitive to parameters, retaining the wild-type phenotype for a wide range of cell signaling rate parameters.

Correlating SHAPE signatures with three-dimensional RNA structures

Selective 2'-hydroxyl acylation analyzed by primer extension (SHAPE) is a facile technique for quantitative analysis of RNA secondary structure. In general, low SHAPE signal values indicate Watson-Crick base-pairing, and high values indicate positions that are single-stranded within the RNA structure. However, the relationship of SHAPE signals to structural properties such as non-Watson-Crick base-pairing or stacking has thus far not been thoroughly investigated. Here, we present results of SHAPE experiments performed on several RNAs with published three-dimensional structures. This strategy allows us to analyze the results in terms of correlations between chemical reactivities and structural properties of the respective nucleotide, such as different types of base-pairing, stacking, and phosphate-backbone interactions. We find that the RNA SHAPE signal is strongly correlated with cis-Watson-Crick/Watson-Crick base-pairing and is to a remarkable degree not dependent on other structural properties with the exception of stacking. We subsequently generated probabilistic models that estimate the likelihood that a residue with a given SHAPE score participates in base-pairing. We show that several models that take SHAPE scores of adjacent residues into account perform better in predicting base-pairing compared with individual SHAPE scores. This underscores the context sensitivity of SHAPE and provides a framework for an improved interpretation of the response of RNA to chemical modification.

Decision making under uncertainty: a neural model based on partially observable markov decision processes

A fundamental problem faced by animals is learning to select actions based on noisy sensory information and incomplete knowledge of the world. It has been suggested that the brain engages in Bayesian inference during perception but how such probabilistic representations are used to select actions has remained unclear. Here we propose a neural model of action selection and decision making based on the theory of partially observable Markov decision processes (POMDPs). Actions are selected based not on a single "optimal" estimate of state but on the posterior distribution over states (the "belief" state). We show how such a model provides a unified framework for explaining experimental results in decision making that involve both information gathering and overt actions. The model utilizes temporal difference (TD) learning for maximizing expected reward. The resulting neural architecture posits an active role for the neocortex in belief computation while ascribing a role to the basal ganglia in belief representation, value computation, and action selection. When applied to the random dots motion discrimination task, model neurons representing belief exhibit responses similar to those of LIP neurons in primate neocortex. The appropriate threshold for switching from information gathering to overt actions emerges naturally during reward maximization. Additionally, the time course of reward prediction error in the model shares similarities with dopaminergic responses in the basal ganglia during the random dots task. For tasks with a deadline, the model learns a decision making strategy that changes with elapsed time, predicting a collapsing decision threshold consistent with some experimental studies. The model provides a new framework for understanding neural decision making and suggests an important role for interactions between the neocortex and the basal ganglia in learning the mapping between probabilistic sensory representations and actions that maximize rewards.

Modeling the fMRI signal via Hierarchical Clustered Hidden Process Models

Machine Learning techniques have been used quite widely for the task of predicting cognitive processes from fMRI data. However, these models do not describe well the fMRI signal when it is generated by multiple cognitive processes that are simultaneously active. In this paper we consider the problem of accurately modeling the fMRI signal of a human subject who is performing a task involving multiple concurrent cognitive processes. We present a Hierarchical Clustering extension of Hidden Process Models which, by taking advantage of automatically discovered similarities in the activation among neighboring voxels, achieves significantly better performance than standard generative models in terms of Average Log Likelihood.

Association chain graphs: modelling etiological pathways

Multiple time-dynamic and interrelated risk factors are usually involved in the complex etiology of disorders. This paper presents a strategy to explore and display visually the relative importance of different association pathways for the onset of disorder over time. The approach is based on graphical chain models, a tool that is powerful but still under-utilized in most fields. Usually, the results of these models are displayed using directed acyclic graphs (DAGs). These draw an edge between a pair of variables whenever the assumption of conditional independence given variables on an earlier or equal temporal footing is violated to a statistically significant extent. In the present paper, the graphs are modified in that confidence intervals for the strengths of associations (statistical main effects) are visualized. These new graphs are called association chain graphs (ACGs). Statistical interactions cause 'edges' between the respective variables within the DAG framework (because the assumption of conditional independence is violated). In contrast they are represented as separate graphs within the subsample where the different association chains may work within the ACG framework. With this new type of graph, more specific information can be displayed whenever the data are essentially described only with statistical main- and two-way interaction effects.

Modeling of activation data in the BrainMap database: detection of outliers

We describe a system for meta-analytical modeling of activation foci from functional neuroimaging studies. Our main vehicle is a set of density models in Talairach space capturing the distribution of activation foci in sets of experiments labeled by lobar anatomy. One important use of such density models is identification of novelty, i.e., low probability database events. We rank the novelty of the outliers and investigate the cause for 21 of the most novel, finding several outliers that are entry and transcription errors or infrequent or non-conforming terminology. We briefly discuss the use of atlases for outlier detection.