Bayesian and deep-learning models applied to the early detection of ovarian cancer using multiple longitudinal biomarkers.

BACKGROUND: Ovarian cancer is the most lethal of all gynecological cancers. Cancer Antigen 125 (CA125) is the best-performing ovarian cancer biomarker which however is still not effective as a screening test in the general population. Recent literature reports additional biomarkers with the potential to improve on CA125 for early detection when using longitudinal multimarker models. METHODS: Our data comprised 180 controls and 44 cases with serum samples sourced from the multimodal arm of UK Collaborative Trial of Ovarian Cancer Screening (UKCTOCS). Our models were based on Bayesian change-point detection and recurrent neural networks. RESULTS: We obtained a significantly higher performance for CA125-HE4 model using both methodologies (AUC 0.971, sensitivity 96.7% and AUC 0.987, sensitivity 96.7%) with respect to CA125 (AUC 0.949, sensitivity 90.8% and AUC 0.953, sensitivity 92.1%) for Bayesian change-point model (BCP) and recurrent neural networks (RNN) approaches, respectively. One year before diagnosis, the CA125-HE4 model also ranked as the best, whereas at 2 years before diagnosis no multimarker model outperformed CA125. CONCLUSIONS: Our study identified and tested different combination of biomarkers using longitudinal multivariable models that outperformed CA125 alone. We showed the potential of multivariable models and candidate biomarkers to increase the detection rate of ovarian cancer.

An Interpretable Machine Learning Method for the Detection of Schizophrenia Using EEG Signals.

In this work we propose a machine learning (ML) method to aid in the diagnosis of schizophrenia using electroencephalograms (EEGs) as input data. The computational algorithm not only yields a proposal of diagnostic but, even more importantly, it provides additional information that admits clinical interpretation. It is based on an ML model called random forest that operates on connectivity metrics extracted from the EEG signals. Specifically, we use measures of generalized partial directed coherence (GPDC) and direct directed transfer function (dDTF) to construct the input features to the ML model. The latter allows the identification of the most performance-wise relevant features which, in turn, provide some insights about EEG signals and frequency bands that are associated with schizophrenia. Our preliminary results on real data show that signals associated with the occipital region seem to play a significant role in the diagnosis of the disease. Moreover, although every frequency band might yield useful information for the diagnosis, the beta and theta (frequency) bands provide features that are ultimately more relevant for the ML classifier that we have implemented.

The Human Body as a Super Network: Digital Methods to Analyze the Propagation of Aging.

Biological aging is a complex process involving multiple biological processes. These can be understood theoretically though considering them as individual networks-e.g., epigenetic networks, cell-cell networks (such as astroglial networks), and population genetics. Mathematical modeling allows the combination of such networks so that they may be studied in unison, to better understand how the so-called "seven pillars of aging" combine and to generate hypothesis for treating aging as a condition at relatively early biological ages. In this review, we consider how recent progression in mathematical modeling can be utilized to investigate aging, particularly in, but not exclusive to, the context of degenerative neuronal disease. We also consider how the latest techniques for generating biomarker models for disease prediction, such as longitudinal analysis and parenclitic analysis can be applied to as both biomarker platforms for aging, as well as to better understand the inescapable condition. This review is written by a highly diverse and multi-disciplinary team of scientists from across the globe and calls for greater collaboration between diverse fields of research.

Multi-Marker Longitudinal Algorithms Incorporating HE4 and CA125 in Ovarian Cancer Screening of Postmenopausal Women.

Longitudinal CA125 algorithms are the current basis of ovarian cancer screening. We report on longitudinal algorithms incorporating multiple markers. In the multimodal arm of United Kingdom Collaborative Trial of Ovarian Cancer Screening (UKCTOCS), 50,640 postmenopausal women underwent annual screening using a serum CA125 longitudinal algorithm. Women (cases) with invasive tubo-ovarian cancer (WHO 2014) following outcome review with stored annual serum samples donated in the 5 years preceding diagnosis were matched 1:1 to controls (no invasive tubo-ovarian cancer) in terms of the number of annual samples and age at randomisation. Blinded samples were assayed for serum human epididymis protein 4 (HE4), CA72-4 and anti-TP53 autoantibodies. Multimarker method of mean trends (MMT) longitudinal algorithms were developed using the assay results and trial CA125 values on the training set and evaluated in the blinded validation set. The study set comprised of 1363 (2-5 per woman) serial samples from 179 cases and 181 controls. In the validation set, area under the curve (AUC) and sensitivity of longitudinal CA125-MMT algorithm were 0.911 (0.871-0.952) and 90.5% (82.5-98.6%). None of the longitudinal multi-marker algorithms (CA125-HE4, CA125-HE4-CA72-4, CA125-HE4-CA72-4-anti-TP53) performed better or improved on lead-time. Our population study suggests that longitudinal HE4, CA72-4, anti-TP53 autoantibodies adds little value to longitudinal serum CA125 as a first-line test in ovarian cancer screening of postmenopausal women.

Monte Carlo method for adaptively estimating the unknown parameters and the dynamic state of chaotic systems.

We propose a Monte Carlo methodology for the joint estimation of unobserved dynamic variables and unknown static parameters in chaotic systems. The technique is sequential, i.e., it updates the variable and parameter estimates recursively as new observations become available, and, hence, suitable for online implementation. We demonstrate the validity of the method by way of two examples. In the first one, we tackle the estimation of all the dynamic variables and one unknown parameter of a five-dimensional nonlinear model using a time series of scalar observations experimentally collected from a chaotic CO2 laser. In the second example, we address the estimation of the two dynamic variables and the phase parameter of a numerical model commonly employed to represent the dynamics of optoelectronic feedback loops designed for chaotic communications over fiber-optic links.

Comparison of Longitudinal CA125 Algorithms as a First-Line Screen for Ovarian Cancer in the General Population.

Purpose: In the United Kingdom Collaborative Trial of Ovarian Cancer Screening (UKCTOCS), women in the multimodal (MMS) arm had a serum CA125 test (first-line), with those at increased risk, having repeat CA125/ultrasound (second-line test). CA125 was interpreted using the "Risk of Ovarian Cancer Algorithm" (ROCA). We report on performance of other serial algorithms and a single CA125 threshold as a first-line screen in the UKCTOCS dataset.Experimental Design: 50,083 post-menopausal women who attended 346,806 MMS screens were randomly split into training and validation sets, following stratification into cases (ovarian/tubal/peritoneal cancers) and controls. The two longitudinal algorithms, a new serial algorithm, method of mean trends (MMT) and the parametric empirical Bayes (PEB) were trained in the training set and tested in the blinded validation set and the performance characteristics, including that of a single CA125 threshold, were compared.Results: The area under receiver operator curve (AUC) was significantly higher (P = 0.01) for MMT (0.921) compared with CA125 single threshold (0.884). At a specificity of 89.5%, sensitivities for MMT [86.5%; 95% confidence interval (CI), 78.4-91.9] and PEB (88.5%; 95% CI, 80.6-93.4) were similar to that reported for ROCA (sensitivity 87.1%; specificity 87.6%; AUC 0.915) and significantly higher than the single CA125 threshold (73.1%; 95% CI, 63.6-80.8).Conclusions: These findings from the largest available serial CA125 dataset in the general population provide definitive evidence that longitudinal algorithms are significantly superior to simple cutoff values for ovarian cancer screening. Use of these newer algorithms requires incorporation into a multimodal strategy. The results highlight the importance of incorporating serial change in biomarker levels in cancer screening/early detection strategies. Clin Cancer Res; 24(19); 4726-33. ©2018 AACR.

A quantitative performance study of two automatic methods for the diagnosis of ovarian cancer.

We present a quantitative study of the performance of two automatic methods for the early detection of ovarian cancer that can exploit longitudinal measurements of multiple biomarkers. The study is carried out for a subset of the data collected in the UK Collaborative Trial of Ovarian Cancer Screening (UKCTOCS). We use statistical analysis techniques, such as the area under the Receiver Operating Characteristic (ROC) curve, for evaluating the performance of two techniques that aim at the classification of subjects as either healthy or suffering from the disease using time-series of multiple biomarkers as inputs. The first method relies on a Bayesian hierarchical model that establishes connections within a set of clinically interpretable parameters. The second technique is a purely discriminative method that employs a recurrent neural network (RNN) for the binary classification of the inputs. For the available dataset, the performance of the two detection schemes is similar (the area under ROC curve is 0.98 for the combination of three biomarkers) and the Bayesian approach has the advantage that its outputs (parameters estimates and their uncertainty) can be further analysed by a clinical expert.

Monte Carlo method for multiparameter estimation in coupled chaotic systems.

We address the problem of estimating multiple parameters of a chaotic dynamical model from the observation of a scalar time series. We assume that the series is produced by a chaotic system with the same functional form as the model, so that synchronization between the two systems can be achieved by an adequate coupling. In this scenario, we propose an efficient Monte Carlo optimization algorithm that iteratively updates the model parameters in order to minimize the synchronization error. As an example, we apply it to jointly estimate the three static parameters of a chaotic Lorenz system.

A comparison of Monte Carlo-based Bayesian parameter estimation methods for stochastic models of genetic networks.

We compare three state-of-the-art Bayesian inference methods for the estimation of the unknown parameters in a stochastic model of a genetic network. In particular, we introduce a stochastic version of the paradigmatic synthetic multicellular clock model proposed by Ullner et al., 2007. By introducing dynamical noise in the model and assuming that the partial observations of the system are contaminated by additive noise, we enable a principled mechanism to represent experimental uncertainties in the synthesis of the multicellular system and pave the way for the design of probabilistic methods for the estimation of any unknowns in the model. Within this setup, we tackle the Bayesian estimation of a subset of the model parameters. Specifically, we compare three Monte Carlo based numerical methods for the approximation of the posterior probability density function of the unknown parameters given a set of partial and noisy observations of the system. The schemes we assess are the particle Metropolis-Hastings (PMH) algorithm, the nonlinear population Monte Carlo (NPMC) method and the approximate Bayesian computation sequential Monte Carlo (ABC-SMC) scheme. We present an extensive numerical simulation study, which shows that while the three techniques can effectively solve the problem there are significant differences both in estimation accuracy and computational efficiency.

Phase control of intermittency in dynamical systems.

We present a nonfeedback method to tame or enhance crisis-induced intermittency in dynamical systems. By adding a small harmonic perturbation to a parameter of the system, the intermittent behavior can be suppressed or enhanced depending on the value of the phase difference between the main driving and the perturbation. The validity of the method is shown both in the model and in an experiment with a CO2 laser. An analysis of this scheme applied to the quadratic map near crisis illustrates the role of phase control in nonlinear dynamical systems.

Adaptive approximation method for joint parameter estimation and identical synchronization of chaotic systems.

We introduce a numerical approximation method for estimating an unknown parameter of a (primary) chaotic system which is partially observed through a scalar time series. Specifically, we show that the recursive minimization of a suitably designed cost function that involves the dynamic state of a fully observed (secondary) system and the observed time series can lead to the identical synchronization of the two systems and the accurate estimation of the unknown parameter. The salient feature of the proposed technique is that the only external input to the secondary system is the unknown parameter which needs to be adjusted. We present numerical examples for the Lorenz system which show how our algorithm can be considerably faster than some previously proposed methods.

Parameter estimation methods for chaotic intercellular networks.

We have investigated simulation-based techniques for parameter estimation in chaotic intercellular networks. The proposed methodology combines a synchronization-based framework for parameter estimation in coupled chaotic systems with some state-of-the-art computational inference methods borrowed from the field of computational statistics. The first method is a stochastic optimization algorithm, known as accelerated random search method, and the other two techniques are based on approximate Bayesian computation. The latter is a general methodology for non-parametric inference that can be applied to practically any system of interest. The first method based on approximate Bayesian computation is a Markov Chain Monte Carlo scheme that generates a series of random parameter realizations for which a low synchronization error is guaranteed. We show that accurate parameter estimates can be obtained by averaging over these realizations. The second ABC-based technique is a Sequential Monte Carlo scheme. The algorithm generates a sequence of "populations", i.e., sets of randomly generated parameter values, where the members of a certain population attain a synchronization error that is lesser than the error attained by members of the previous population. Again, we show that accurate estimates can be obtained by averaging over the parameter values in the last population of the sequence. We have analysed how effective these methods are from a computational perspective. For the numerical simulations we have considered a network that consists of two modified repressilators with identical parameters, coupled by the fast diffusion of the autoinducer across the cell membranes.