Demixed principal component analysis of neural population data

  1. Dmitry Kobak  Is a corresponding author
  2. Wieland Brendel
  3. Christos Constantinidis
  4. Claudia E Feierstein
  5. Adam Kepecs
  6. Zachary F Mainen
  7. Ranulfo Romo
  8. Xue-Lian Qi
  9. Naoshige Uchida
  10. Christian K Machens
  1. Champalimaud Centre for the Unknown, Portugal
  2. Wake Forest University School of Medicine, United States
  3. Cold Spring Harbor Laboratory, United States
  4. Universidad Nacional Autónoma de México, Mexico
  5. Harvard University, United States

Abstract

Neurons in higher cortical areas, such as the prefrontal cortex, are often tuned to a variety of sensory and motor variables, and are therefore said to display mixed selectivity. This complexity of single neuron responses can obscure what information these areas represent and how it is represented. Here we demonstrate the advantages of a new dimensionality reduction technique, demixed principal component analysis (dPCA), that decomposes population activity into a few components. In addition to systematically capturing the majority of the variance of the data, dPCA also exposes the dependence of the neural representation on task parameters such as stimuli, decisions, or rewards. To illustrate our method we reanalyze population data from four datasets comprising different species, different cortical areas and different experimental tasks. In each case, dPCA provides a concise way of visualizing the data that summarizes the task-dependent features of the population response in a single figure.

Article and author information

Author details

  1. Dmitry Kobak

    Champalimaud Neuroscience Program, Champalimaud Centre for the Unknown, Lisbon, Portugal
    For correspondence
    dmitry.kobak@neuro.fchampalimaud.org
    Competing interests
    No competing interests declared.
  2. Wieland Brendel

    Champalimaud Neuroscience Program, Champalimaud Centre for the Unknown, Lisbon, Portugal
    Competing interests
    No competing interests declared.
  3. Christos Constantinidis

    Wake Forest University School of Medicine, Winston-Salem, United States
    Competing interests
    No competing interests declared.
  4. Claudia E Feierstein

    Champalimaud Neuroscience Program, Champalimaud Centre for the Unknown, Lisbon, Portugal
    Competing interests
    No competing interests declared.
  5. Adam Kepecs

    Cold Spring Harbor Laboratory, Cold Spring Harbor, United States
    Competing interests
    No competing interests declared.
  6. Zachary F Mainen

    Champalimaud Neuroscience Program, Champalimaud Centre for the Unknown, Lisbon, Portugal
    Competing interests
    No competing interests declared.
  7. Ranulfo Romo

    Instituto de Fisiología Celular-Neurociencias, Universidad Nacional Autónoma de México, Ciudad de México, Mexico
    Competing interests
    No competing interests declared.
  8. Xue-Lian Qi

    Wake Forest University School of Medicine, Winston-Salem, United States
    Competing interests
    No competing interests declared.
  9. Naoshige Uchida

    Harvard University, Cambridge, United States
    Competing interests
    Naoshige Uchida, Reviewing editor, eLife.
  10. Christian K Machens

    Champalimaud Neuroscience Program, Champalimaud Centre for the Unknown, Lisbon, Portugal
    Competing interests
    No competing interests declared.

Copyright

© 2016, Kobak et al.

This article is distributed under the terms of the Creative Commons Attribution License permitting unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 38,691
    views
  • 6,664
    downloads
  • 448
    citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Dmitry Kobak
  2. Wieland Brendel
  3. Christos Constantinidis
  4. Claudia E Feierstein
  5. Adam Kepecs
  6. Zachary F Mainen
  7. Ranulfo Romo
  8. Xue-Lian Qi
  9. Naoshige Uchida
  10. Christian K Machens
(2016)
Demixed principal component analysis of neural population data
eLife 5:e10989.
https://doi.org/10.7554/eLife.10989

Share this article

https://doi.org/10.7554/eLife.10989

Further reading

    1. Neuroscience
    John P Grogan, Matthias Raemaekers ... Sanjay G Manohar
    Research Article

    Motivation depends on dopamine, but might be modulated by acetylcholine which influences dopamine release in the striatum, and amplifies motivation in animal studies. A corresponding effect in humans would be important clinically, since anticholinergic drugs are frequently used in Parkinson’s disease, a condition that can also disrupt motivation. Reward and dopamine make us more ready to respond, as indexed by reaction times (RT), and move faster, sometimes termed vigour. These effects may be controlled by preparatory processes that can be tracked using electroencephalography (EEG). We measured vigour in a placebo-controlled, double-blinded study of trihexyphenidyl (THP), a muscarinic antagonist, with an incentivised eye movement task and EEG. Participants responded faster and with greater vigour when incentives were high, but THP blunted these motivational effects, suggesting that muscarinic receptors facilitate invigoration by reward. Preparatory EEG build-up (contingent negative variation [CNV]) was strengthened by high incentives and by muscarinic blockade, although THP reduced the incentive effect. The amplitude of preparatory activity predicted both vigour and RT, although over distinct scalp regions; frontal activity predicted vigour, whereas a larger, earlier, central component predicted RT. The incentivisation of RT was partly mediated by the CNV, though vigour was not. Moreover, the CNV mediated the drug’s effect on dampening incentives, suggesting that muscarinic receptors underlie the motivational influence on this preparatory activity. Taken together, these findings show that a muscarinic blocker impairs motivated action in healthy people, and that medial frontal preparatory neural activity mediates this for RT.

    1. Neuroscience
    Samyogita Hardikar, Bronte Mckeown ... Jonathan Smallwood
    Research Article

    Complex macro-scale patterns of brain activity that emerge during periods of wakeful rest provide insight into the organisation of neural function, how these differentiate individuals based on their traits, and the neural basis of different types of self-generated thoughts. Although brain activity during wakeful rest is valuable for understanding important features of human cognition, its unconstrained nature makes it difficult to disentangle neural features related to personality traits from those related to the thoughts occurring at rest. Our study builds on recent perspectives from work on ongoing conscious thought that highlight the interactions between three brain networks – ventral and dorsal attention networks, as well as the default mode network. We combined measures of personality with state-of-the-art indices of ongoing thoughts at rest and brain imaging analysis and explored whether this ‘tri-partite’ view can provide a framework within which to understand the contribution of states and traits to observed patterns of neural activity at rest. To capture macro-scale relationships between different brain systems, we calculated cortical gradients to describe brain organisation in a low-dimensional space. Our analysis established that for more introverted individuals, regions of the ventral attention network were functionally more aligned to regions of the somatomotor system and the default mode network. At the same time, a pattern of detailed self-generated thought was associated with a decoupling of regions of dorsal attention from regions in the default mode network. Our study, therefore, establishes that interactions between attention systems and the default mode network are important influences on ongoing thought at rest and highlights the value of integrating contemporary perspectives on conscious experience when understanding patterns of brain activity at rest.