Sensory cortex is optimised for prediction of future input
Abstract
Neurons in sensory cortex are tuned to diverse features in natural scenes. But what determines which features neurons become selective to? Here we explore the idea that neuronal selectivity is optimised to represent features in the recent sensory past that best predict immediate future inputs. We tested this hypothesis using simple feedforward neural networks, which were trained to predict the next few video or audio frames in clips of natural scenes. The networks developed receptive fields that closely matched those of real cortical neurons in different mammalian species, including the oriented spatial tuning of primary visual cortex, the frequency selectivity of primary auditory cortex and, most notably, their temporal tuning properties. Furthermore, the better a network predicted future inputs the more closely its receptive fields resembled those in the brain. This suggests that sensory processing is optimised to extract those features with the most capacity to predict future input.
Data availability
All custom code used in this study was implemented in MATLAB and Python. We have uploaded the code to a public Github repository. The raw auditory experimental data is available at https://osf.io/ayw2p/. The movies and sounds used for training the models are all publicly available at the websites detailed in the Methods.
Article and author information
Author details
Funding
Clarendon Fund
- Yosef Singer
- Yayoi Teramoto
University Of Oxford
- Nicol S Harper
Action on Hearing Loss (PA07)
- Nicol S Harper
Biotechnology and Biological Sciences Research Council (BB/H008608/1)
- Nicol S Harper
Wellcome (WT10525/Z/14/Z)
- Yayoi Teramoto
Wellcome (WT076508AIA)
- Ben DB Willmore
Wellcome (WT108369/Z/2015/Z)
- Ben DB Willmore
Wellcome (WT076508AIA)
- Andrew J King
Wellcome (WT108369/Z/2015/Z)
- Andrew J King
Wellcome (WT082692)
- Nicol S Harper
Wellcome (WT076508AIA)
- Nicol S Harper
Wellcome (WT108369/Z/2015/Z)
- Nicol S Harper
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
Reviewing Editor
- Jack L Gallant, University of California, Berkeley, United States
Ethics
Animal experimentation: Auditory RFs of neurons were recorded in the primary auditory cortex (A1) and anterior auditory field (AAF) of 5 pigmented ferrets of both sexes (all > 6 months of age) and used as a basis for comparison with the RFs of model units trained on auditory stimuli. These recordings were performed under license from the UK Home Office and were approved by the University of Oxford Committee on Animal Care and Ethical Review. Full details of the recording methods are described in earlier studies [45,90]. Briefly, we induced general anaesthesia with a single intramuscular dose of medetomidine (0.022 mg · kg−1 · h−1) and ketamine (5 mg · kg−1 · h−1), which was then maintained with a continuous intravenous infusion of medetomidine and ketamine in saline. Oxygen was supplemented with a ventilator, and we monitored vital signs (body temperature, end-tidal CO2, and the electrocardiogram) throughout the experiment. The temporal muscles were retracted, a head holder was secured to the skull surface, and a craniotomy and a durotomy were made over the auditory cortex. Extracellular recordings were made using silicon probe electrodes (Neuronexus Technologies) and acoustic stimuli were presented via Panasonic RPHV27 earphones, which were coupled to otoscope specula that were inserted into each ear canal, and driven by Tucker-Davis Technologies System III hardware (48 kHz sample rate).
Version history
- Received: October 4, 2017
- Accepted: June 16, 2018
- Accepted Manuscript published: June 18, 2018 (version 1)
- Accepted Manuscript updated: June 22, 2018 (version 2)
- Version of Record published: August 24, 2018 (version 3)
- Version of Record updated: September 18, 2018 (version 4)
Copyright
© 2018, Singer 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
-
- 7,605
- views
-
- 1,123
- downloads
-
- 51
- citations
Views, downloads and citations are aggregated across all versions of this paper published by eLife.
Download links
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)
Further reading
-
- Medicine
- Neuroscience
Retinitis pigmentosa (RP) is an inherited retinal disease in which there is a loss of cone-mediated daylight vision. As there are >100 disease genes, our goal is to preserve cone vision in a disease gene-agnostic manner. Previously we showed that overexpressing TXNIP, an α-arrestin protein, prolonged cone vision in RP mouse models, using an AAV to express it only in cones. Here, we expressed different alleles of Txnip in the retinal pigmented epithelium (RPE), a support layer for cones. Our goal was to learn more of TXNIP’s structure-function relationships for cone survival, as well as determine the optimal cell type expression pattern for cone survival. The C-terminal half of TXNIP was found to be sufficient to remove GLUT1 from the cell surface, and improved RP cone survival, when expressed in the RPE, but not in cones. Knock-down of HSP90AB1, a TXNIP-interactor which regulates metabolism, improved the survival of cones alone and was additive for cone survival when combined with TXNIP. From these and other results, it is likely that TXNIP interacts with several proteins in the RPE to indirectly support cone survival, with some of these interactions different from those that lead to cone survival when expressed only in cones.
-
- Neuroscience
Vesicles within presynaptic terminals are thought to be segregated into a variety of readily releasable and reserve pools. The nature of the pools and trafficking between them is not well understood, but pools that are slow to mobilize when synapses are active are often assumed to feed pools that are mobilized more quickly, in a series. However, electrophysiological studies of synaptic transmission have suggested instead a parallel organization where vesicles within slowly and quickly mobilized reserve pools would separately feed independent reluctant- and fast-releasing subdivisions of the readily releasable pool. Here, we use FM-dyes to confirm the existence of multiple reserve pools at hippocampal synapses and a parallel organization that prevents intermixing between the pools, even when stimulation is intense enough to drive exocytosis at the maximum rate. The experiments additionally demonstrate extensive heterogeneity among synapses in the relative sizes of the slowly and quickly mobilized reserve pools, which suggests equivalent heterogeneity in the numbers of reluctant and fast-releasing readily releasable vesicles that may be relevant for understanding information processing and storage.