Tracking the topology of neural manifolds across populations

• 成果类型: Article
• 署名作者: Yoon, Iris H. R.; Henselman-Petrusek, Gregory; Yu, Yiyi; Ghrist, Robert; Smith, Spencer LaVere; Giusti, Chad
• 署名单位: Wesleyan University; United States Department of Energy (DOE); Pacific Northwest National Laboratory; University of California System; University of California Santa Barbara; University of Pennsylvania; University of Pennsylvania; Oregon State University
• 刊物名称: PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA
• ISSN/ISSBN: 0027-9045
• DOI: 10.1073/pnas.2407997121
• 发表日期: 2024-11-12
• 关键词: functional specialization; receptive-fields; visual-cortex; spatial map; propagation; projections; direction; cells

摘要

Neural manifolds summarize the intrinsic structure of the information encoded bya population of neurons. Advances in experimental techniques have made simulta-neous recordings from multiple brain regions increasingly commonplace, raising thepossibility of studying how these manifolds relate across populations. However, whenthe manifolds are nonlinear and possibly code for multiple unknown variables, it ischallenging to extract robust and falsifiable information about their relationships.We introduce a framework, called the method of analogous cycles, for matchingtopological features of neural manifolds using only observed dissimilarity matriceswithin and between neural populations. We demonstrate via analysis of simulationsand in vivo experimental data that this method can be used to correctly identify multipleshared circular coordinate systems across both stimuli and inferred neural manifolds.Conversely, the method rejects matching features that are not intrinsic to one of thesystems. Further, as this method is deterministic and does not rely on dimensionalityreduction or optimization methods, it is amenable to direct mathematical investigationand interpretation in terms of the underlying neural activity. We thus propose themethod of analogous cycles as a suitable foundation for a theory of cross-populationanalysis via neural manifolds.