Control of Dynamics in Brain Networks Evelyn Tang Department of Bioengineering, University of Pennsylvania, PA 19104 Danielle S. Bassett Department of Bioengineering, Department of Physics & Astronomy, Department of Neurology, Department of Electrical & Systems Engineering, University of Pennsylvania, PA 19104 (Dated: May 16, 2018) The ability to effectively control brain dynamics holds great promise for the enhancement of cognitive function in humans, and the betterment of their quality of life. Yet, success- fully controlling dynamics in neural systems is challenging, in part due to the immense complexity of the brain and the large set of interactions that can drive any single change. While we have gained some understanding of the control of single neurons, the control of large-scale neural systems|networks of multiply interacting components|remains poorly understood. Efforts to address this gap include the construction of tools for the control of brain networks, mostly adapted from control and dynamical systems theory. Informed by current opportunities for practical intervention, these theoretical contri- butions provide models that draw from a wide array of mathematical approaches. We present recent developments for effective strategies of control in dynamic brain networks, and we also describe potential mechanisms that underlie such processes. We review ef- forts in the control of general neurophysiological processes with implications for brain development and cognitive function, as well as the control of altered neurophysiological processes in medical contexts such as anesthesia administration, seizure suppression, and deep-brain stimulation for Parkinson's disease. We conclude with a forward-looking discussion regarding how emerging results from network control|especially approaches that deal with nonlinear dynamics or more realistic trajectories for control transitions| could be used to directly address pressing questions in neuroscience. CONTENTS 1. Dynamical characteristics and clinical relevance 13 2. Structural drivers of synchrony: Graph I. Introduction 2 architecture and symmetries 13 B. The cost of controlling specific trajectories 14 II. How does the brain control itself? 3 C. Empirical tools for control of specific neural dynamics or pathways 16 III. Network control theory 4 A. Control of linear dynamics 4 VII. Emerging control methods with potential utility in B. Key driver nodes 4 neuroscience 16 C. Control energy and metrics 5 A. Broader control regimes 16 D. Application to brain networks 6 1. Nonlinear dynamics 16 2. Time-dependent control 17 IV. Understanding healthy brain function through control 3. Realistic control trajectories 17 arXiv:1701.01531v3 [q-bio.QM] 14 May 2018 theory 7 B. Exploiting system properties 18 A. Network control as a partial mechanism for cognitive 1. Compensatory perturbations or noise 18 control 7 2. Network topology 18 B. Network control and cognitive performance 8 C. Evolution of network control in development 9 VIII. Conclusion 19 D. Open questions in control and cognition 9 Acknowledgments 19 V. Targeting therapeutic interventions to maximize References 19 beneficial outcomes to patients 10 A. Anesthesia titration 10 B. Deep-brain stimulation for Parkinson's disease 10 C. Non-invasive transcranial stimulation 11 D. Seizure suppression in epilepsy 12 VI. Control of specific neural dynamics or pathways 13 A. Synchrony of neural populations 13 2 I. INTRODUCTION (Fornito and Bullmore, 2015; Sporns et al., 2005). Here we use the term \network" in the sense that is common The brain displays a wealth of complex dynamics in network science (Newman, 2010). A brain network is across various spatial and temporal scales (Betzel and a graph whose nodes represent units of the brain that Bassett, 2016; Kopell et al., 2014). From 302 neurons perform a specific function, like vision or audition (Bull- in the nematode worm C. elegans (Bentley et al., 2016; more and Bassett, 2011). At the large-scale, these units Varier and Kaiser, 2011) to some 86 billion neurons in may be several centimeters of tissue, while at the small the adult human (von Bartheld et al., 2016; Herculano- scale, these units may be individual neurons. In struc- Houzel et al., 2007), the units that drive brain func- tural brain graphs, the edges can represent structural tion are large in their number but even more compli- links such as fiber bundles at the large scale (Bassett cated in their interactions. Far from the canonical mod- et al., 2011; Hagmann et al., 2008) or synapses at the els in statistical mechanics stemming from either crys- small scale. In functional brain graphs, the edges rep- talline or random structure, the brain displays a hetero- resent synchronized dynamics that form functional links geneous pattern of interconnections (Bassett and Bull- (Achard et al., 2006; Stam, 2004) between these units. more, 2016; Castellana and Bialek, 2014; Fraiman et al., While both structural and functional links can be mea- 2009) that fundamentally constrains the propagation of sured directly from structural and functional data, re- activity. Understanding these dynamics remains of pri- spectively, extensive efforts have also sought to address mary interest in the field of neurophysics (Gao and Gan- the questions of (i) whether structural topology can be guli, 2015; Scott, 1977). An underlying assumption of inferred from functional traces (using, for example, struc- these investigations is that such dynamics or observed tural equation modeling), and (ii) whether functional neural activity can contain structure that forms represen- traces can be inferred from structural linkage patterns tations about incoming stimuli or underlying neural pro- (using, for example, neural mass models). Throughout cesses. An emerging and increasingly tractable avenue for this exposition, we will assume that structural links have understanding the mechanisms of these dynamics lies in been directly measured, rather than inferred. the notion of control, or how to effectively guide neural The use of the network formalism to probe brain dy- dynamics. How are brain dynamics controlled intrinsi- namics has a rich and pervasive heritage in seminal work cally in the awake, behaving animal? Can we harness at the intersection between physics and neuroscience. natural principles of control in neural systems to better One particularly impactful contribution was that of Hop- guide therapeutic interventions? field, who successfully connected dynamical processes to The increase in available experimental neurotechnolo- neural representations in an Ising model (Hopfield, 1982). gies (Chang, 2015; Nag and Thakor, 2016; Patil and States that minimized the energy function formed dy- Thakor, 2016), as well as more sophisticated compu- namical attractors and representations of memory. This tational tools (Glaser and Kording, 2016; Marblestone early contribution was extended and formalized by Amit et al., 2016) and theoretical models (Giusti et al., 2016), et al. (1985) and Gardner (1987), clearly demonstrat- has recently made it possible to tackle these questions ing the power of interacting networks in the modeling from fundamentally new angles. While at present there of complex neural processes. Here we expand the link is no comprehensive theory of control in the brain that between physics and neuroscience in the context of the we can refer to, the pursuit of such a theory remains network formalism by focusing on the control of brain critically important, having implications for our under- networks, enabling us to build a theoretical understand- standing of healthy neurophysiological processes, and our ing regarding biological processes and associated dynam- ability to intervene when those healthy processes go awry ics that occur across spatially distributed neural systems. in neurological disease and psychiatric disorders (Bassett In addition, strategies for intervention and control tar- and Khambhati, 2017; Chen et al., 2014; Johnson et al., gets can be designed through modeling dynamics in net- 2013). Several recent models propose new ways to control works of neurons or brain regions. Should the reader neural activity and neural rhythms, and further provide instead be searching for an excellent treatment of var- mechanistic insights into the rules by which brain dy- ious control methods for single neurons or for ensem- namics are (and can be) guided. Hence, it is timely to bles of neurons, we direct them to the recent textbook discuss these emerging developments, and to seek to tie by Schiff (2012), and to references therein. For further them together into a meaningful theoretical field that can details on emerging control technologies in the brain| be used to tackle current open questions in neuroscience especially invasive electrical and optical stimulation at and medicine. rapid timescales (milliseconds or below)|and associated Motivated by recent progress in understanding brain modelling approaches, please see the recent review by function from the perspective of interacting networks Ritt and Ching (2015). (Bassett and Bullmore, 2006, 2009; Bullmore and Sporns, The remainder of this Colloquium is organized as fol- 2012; Kaiser, 2011), we focus on systems-level control of lows. In Sec. II we draw inspiration for understanding either local neural circuits or whole-brain connectomes control of brain networks by considering how the brain 3 itself enacts intrinsic control. In particular, we briefly by proponents of connectionism (Medler, 1998), which discuss important