Patrick Cahill

Brain networks exhibit non-trivial modular organization, with groups of densely connected areas participating in specialized functions. Traditional community detection algorithms assign each node to one module, but this representation cannot capture integrative, multifunctional nodes that span multiple communities. Despite the increasing availability of overlapping community detection algorithms (OCDAs) to capture such integrative nodes, there is no objective procedure for selecting the most appropriate method and its parameters for a given problem. Here, we overcome this limitation by introducing a data-driven method for selecting an OCDA and its parameters from performance on a tailored ensemble of generated benchmark networks, assessing 22 unique algorithms and parameter settings. Applied to the human right-hemisphere structural connectome, we find that the “order statistics local optimization method” (OSLOM) best identifies ground-truth overlapping structure in the benchmark ensemble, yielding a seven-network decomposition of the right-hemisphere cortex. These modules are bridged by 15 overlapping regions that generally sit at the apex of the putative cortical hierarchy—suggesting integrative, higher order function—with network participation increasing along the cortical hierarchy, a finding not supported using a non-overlapping modular decomposition. This data-driven approach to selecting OCDAs is applicable across domains, opening new avenues to detecting and quantifying informative structures in complex real-world networks.

The interdisciplinary time-series analysis literature encompasses thousands of statistical features for quantifying interpretable properties of dynamical data. For any given application, however, it is likely that only a small subset of informative time-series features is required to capture the dynamical quantities of interest. While comprehensive libraries of time-series features have been developed, it is therefore useful to construct reduced and computationally efficient subsets for specific applications. In this work, we demonstrate a systematic process to deduce such a reduced set, focusing on the problem of distinguishing changes to functional magnetic resonance imaging (fMRI) time series induced by various experimental manipulations to excitatory and inhibitory neural activity in mouse cortical circuits. We reduce a comprehensive library of over 7000 candidate time-series features down to a subset of 16 features, which we call catchaMouse16, that aims to both: (i) accurately characterize biologically relevant properties of fMRI time series; and (ii) minimize inter-feature redundancy. The catchaMouse16 feature set accurately classifies experimental perturbations of neuronal activity from fMRI recordings and generalizes well to new mouse and human resting-state fMRI datasets, where it tracks spatial variations in excitatory and inhibitory cortical cell densities—often with greater statistical power than the full hctsa feature set. We provide an efficient, open-source implementation of the catchaMouse16 feature set in C (achieving an approximately 60 times speed-up relative to the native Matlab code of the same features), with wrappers for Python and Matlab. This work demonstrates a procedure to reduce a large candidate time-series feature set down to the key statistical properties inherent to mouse fMRI dynamics, which can in turn be used to efficiently quantify and interpret informative dynamical patterns in neural time series.

Background: While osteoarticular infections of the bones and joints in children usually resolve completely with adequate treatment, some children will develop complications. Predicting which children are at highest risk of complications might enable prevention through more aggressive treatment. Methods: We used mutual information and Kullback-Leibler divergence methods to compare the characteristics of osteoarticular infections at the same tertiary institution in Australia in two time periods, 2002-2007 (N=295) and 2016-2018 (N=192). We used expert knowledge to develop a causal directed acyclic graph (DAG) that depicts the mechanistic pathways of osteoarticular infections and their progression to complications. Guided by the DAG, we developed three logistic regression (LR) models for predicting complications and evaluated their area under curve (AUC) to assess their performance for predicting complication. Results: We observed a shift in the approach to diagnostic testing over the two time periods, with an increase in the number of blood cultures performed and a decrease in the rate of wound cultures. Fourteen of 43 test PCR tests (33%) for K. kingae recorded positive results. The established causal DAG clarified how the underlying, latent and dynamic biological processes become manifest as data. Utilising only data available at the initial point of care, the best LR model identified an optimal feature set that achieved an AUC of 0.85 for predicting complications. Conclusions: Supported by domain expert knowledge and data, causal and statistical approaches were combined to offer valuable insights for predicting progression to complicated disease for children with osteoarticular infections. Key messages: This study facilitates the understanding of change in epidemiology and clinical management of paediatric osteomyelitis and septic arthritis by examining a prospective cohort of 295 cases (2016-2018) in comparison to a similar study collected at an earlier date (2002-2007) from the same facility. We present a clarified understanding of the mechanistic pathways of osteoarticular infections and their progression to complications through the development of a causal directed acyclic graph (DAG) based on previously published DAG, recently collected data and domain expert knowledge. We demonstrate an approach to combine causal DAG, BN-based predictions of the causative pathogen, logistic regression modelling, and recursive feature elimination to achieve optimal performance in predicting complicated disease.

The interdisciplinary time-series analysis literature encompasses thousands of statistical features for quantifying interpretable properties of dynamical data. But for any given application, it is likely that just a small subset of informative time-series features is required to capture the dynamical quantities of interest. So, while comprehensive libraries of time-series features have been developed, it is useful to construct reduced and computationally efficient subsets for specific applications. In this work, we demonstrate a systematic process to deduce such a reduced set, focused on the problem of distinguishing changes to functional Magnetic Resonance Imaging (fMRI) time series caused by a range of experimental manipulations of excitatory and inhibitory neural activity in mouse cortical circuits. We reduce a comprehensive library of over 7000 candidate time-series features down to a subset of 16 features, which we call catchaMouse16 , that aims to both: (i) accurately characterize biologically relevant properties of fMRI time series; and (ii) minimize inter-feature redundancy. The catchaMouse16 feature set accurately classifies experimental perturbations of neuronal activity from fMRI recordings, and also shows strong generalization performance on an unseen mouse and human resting-state fMRI data where it tracks spatial variations in excitatory and inhibitory cortical cell densities, often with greater statistical power than the full hctsa feature set. We provide an efficient, open-source implementation of the catchaMouse16 feature set in C (achieving an approximately 60 times speed-up relative to the native Matlab code of the same features), with wrappers for Python and Matlab. This work demonstrates a procedure to reduce a large candidate time-series feature set down to the key statistical properties of mouse fMRI dynamics that can be used to efficiently quantify and interpret informative dynamical patterns in neural time series.