kordinglab/spykes

Tools for spike data analysis and visualization

spykes

Almost any electrophysiology study of neural spiking data relies on a battery of standard analyses. Raster plots and peri-stimulus time histograms aligned to stimuli and behavior provide a snapshot visual description of neural activity. Similarly, tuning curves are the most standard way to characterize how neurons encode stimuli or behavioral preferences. With increasing popularity of population recordings, maximum-likelihood decoders based on tuning models are becoming part of this standard.

Yet, virtually every lab relies on a set of in-house analysis scripts to go from raw data to summaries. We want to improve this status quo in order to enable easier sharing, better reproducibility and fewer bugs.

Spykes is a collection of Python tools to make the visualization and analysis of neural data easy and reproducible.

For more, see the documentation.

Installation

Spykes can be installed using

pip install spykes

For more detailed installation options, see the documentation.

Authors

• Pavan Ramkumar
• Hugo Fernandes

Acknowledgments

• Konrad Kording for funding and support

spykes's Issues

make the fitted tuning curve continuous

demonstrate neuropop using center out reaching data

move slow_exp into a numba function

to speed up many 1000's of calls made to the same function

style is not working for the first neuropop display panel

unless it is loaded in the beginning of the notebook

separate events from features in all notebook examples

Improvements to example notebooks

List of possible improvements for example notebooks:

reaching_example:

• color code by reaching direction the circular plots in decoding from pop

simul_example:

• use new simulated spike data (in a restricted angular direction) for evaluation of decoding performance and plotting

demonstrate psth using center our reach data

score function for NeuroPop

• goodness of fit measure for tuning curves fit and for decoding

add legend to PSTHs

solve time problem in PSTH

onset is slightly shifted, maybe a bin off

translate for loops into matrix/ vector multiplications in fit() and decode()

spykes.mplstyle not found on import

from neurovis import NeuroVis

---------------------------------------------------------------------------
IOError                                   Traceback (most recent call last)
<ipython-input-11-c37dd3c942ac> in <module>()
----> 1 from neurovis import NeuroVis

/usr/local/home/jkiggins/Code/spykes/neurovis/__init__.py in <module>()
----> 1 from .neurovis import NeuroVis
      2 __version__ = '0.1.dev'

/usr/local/home/jkiggins/Code/spykes/neurovis/neurovis.py in <module>()
      2 import matplotlib.pyplot as plt
      3 
----> 4 plt.style.use('../mpl_styles/spykes.mplstyle')
      5 
      6 

/home/jkiggins/envs/jkiggins/lib/python2.7/site-packages/matplotlib/style/core.pyc in use(style)
     88                        "not a valid URL or path. See `style.available` for "
     89                        "list of available styles.")
---> 90                 raise IOError(msg % style)
     91 
     92 

IOError: '../mpl_styles/spykes.mplstyle' not found in the style library and input is not a valid URL or path. See `style.available` for list of available styles.

fix numba dependency

The from numba.decorators import autojitin neuropop.py makes an error if the user doesn't have numba installed

The README states its not a dependency but rather for optimization.

Thus, two potential solutions are

1. It should be considered a dependency and reflect that in the README
2. try importing it if it exists (see: http://stackoverflow.com/questions/5847934/how-to-check-if-python-module-exists-and-can-be-imported)

fix poisson noise bug in neuropop simulate

demonstrate get_spikecounts and get_trialtype functions of NeuroVis

create simple docs page providing documentation for callable methods

add rasters to reaching example

plot PSTHs of population as rasters/ heat maps

Suggestion from Lee Miller's Beer & Brainstorm:

"We often plot mean PSTHs of neurons in a population as a raster or heat map. We may sort these by

• preferred stimulus
• firing rate
• peak latency of PSTH"

Takeaway:

Write a new class called PopVis() which

• takes a population of spike counts
• computes preferred stimuli
• computes PSTHs
• plots them as a heat map (rows = neurons, cols = time)
• takes a sortby argument that sorts neurons by one of above three ways

allow for psth and rasters to be organized into subplots

fix bug for empty spiketimes in get_spikecounts

separate example notebooks from tutorial notebooks

fix poor decoding performance in reaching dataset

Change unit of time window in get_spikecounts to milliseconds

display routine for NeuroPop

plot spikecounts vs. feature and fitted tuning curve (give color option)

use a dataset from crcns.org

update readme to include dependencies

define a uniform and callable plotting style

add vanilla von mises method for tuning curve fit and decoding in NeuroPop

Currently, the $$b + g*\exp(\cos(\theta - \theta_0))$$ model works, but it's non-convex and very slow.

I think we should add a simple $$\exp(b_0 + b_cos_\cos(\theta) + b_sin_\sin(\theta))$$ model for tunefit(), encode() and decode().

We can give the user the option of choosing the model they prefer

make sta method in NEURON class

include faster non-rigorous way of computing rasters as a non-default option

it will be good to have this option for faster data exploration when the number of event_times is large

make decoder visualization polar/ circular

The readme file sucks.

Konrad, can you take a look? It needs some love.

Neuropop utility for population summary plots

• population summary of fitted preferred features (histogram)
• population summary of fitted preferred features vs. neuron

bootstrap CI for estimated parameters

store bootstrap CIs of params if requested in a returnable called stats

find better solution for the style file path problem

now it works with pip -e but not with python setup.py install

make setup.py for package installation

implement sortby and sortorder for plot_raster in NeuroVis

Currently, we have sort which is a bool and sorts by mean spike rate. Let's do the same thing as we did in PopVis() and implement sortby and sortorder here too.

We can do this once issue #60 is solved.

improving precision of event rasters

The method of generating the rasters strikes me as incredibly imprecise. Please correct me if my understanding of the algorithm is wrong.

We create a single bins vector that spans the spiketimes, then...

# bin the spikes into time bins
spike_counts = np.histogram(self.spiketimes, bins)[0]

# bin the events into time bins
event_counts = np.histogram(selected_events, bins)[0]
event_bins = np.where(event_counts > 0)[0]

raster = np.array([(spike_counts[(i+window[0]/binsize): \
                                 (i+window[1]/binsize+1)]) \
                   for i in event_bins])

Spikes AND events are binned in this vector and the raster is built from the counts aligned to the bins that contain events.

This approach means that any spike that lands in the same bin as the event is counted in that bin, no matter whether it occurred before the event of after.

Further, if events are not regular, then sometimes the event will happen early in the bin and the bin will consist mostly of spikes that happen after the event and other times, visa versa.

I expected that a unique bins vector would be constructed for each event, spanning the window so that spikes were clearly assigned before or after the event.

improve docstring for psth

simulate method for NeuroPop

make plot_psth method in Spyke class

the first time pop.display is ran the style doesn’t seem to be loaded

only for the second time, see first panel here:

Trial-varying firing rate estimation algorithm

Suggestion by Vincent Ferrera at Columbia:

Paper
Matlab Code

Thanks @belarius ! Welcome to submit PR

translate ipynb to rst for docs

fix error in psth and rasters in low firing neurons or conditions with zero trials

add travis for CI

add circle CI for docs

upload mini dataset for examples to Repo

suggested improvements to the API

This is a somewhat long feature request.

I'm really excited about these packages and I think it's a major step in the right direction for getting plotting functions of neural time-series data. However, I think the current API is severely limited, especially with respect to the features and conditions arguments.

I would love to see an API which bears more similarity to @mwaskom's seaborn.

Seaborn's plotting functionality is amazing. For pretty much all of seaborn's plotting functions, you pass in a pandas dataframe as well as the column names you want plotted against each other. Then, you can split the data into colors by passing a column name for the hue argument, or break out the plot into an array of subplots where each col/row is defined according to a different factor.

The utility of this approach is more clear looking at the factorplot examples: https://stanford.edu/~mwaskom/software/seaborn/generated/seaborn.factorplot.html

Since I use seaborn extensively for data exploration already, wrote up a few functions that wrap the neurovis plotting functions in something more seaborn-like so you can see what I mean. Here's a notebook that demonstrates (a) getting the features and conditions into a friendly pandas dataframe and (b) plotting with seaborn style.

https://gist.github.com/neuromusic/f6fe2c5d7c1101811b5cd497b98a516c

Currently, by treating features and conditions as separate arguments, neurovis is incredibly limited. For one, the dictionary format of the conditions is not intuitive. Second, it's hard to compute complex conditions from the dictionary format that is currently used -- it seems largely limited to performing "AND" operations between factors and only filtering factors with ranges. It is better, I think, to let the user use pandas to construct factors with whatever logic they want.

In my above example, I only used "hue" since neurovis already supports this. Along with fixing #22, it would not be too hard to add a col or row factor. Then, in my example in the gist above, one could quickly plot the same plot as above but breaking out plots for correct and incorrect trials into rows like so...

trials = events_df[(events_df['GoodTrials']==True)]

# plot the PSTH
psthplot('RealCueTimes',trials,
         spike_times,nid=neuron_n,
         hue='response',
         row='correct',
         window=window,
         binsize=100)

One could also then also address #42 and/or other PSTH-like computations (e.g. gaussian kernels, poisson GLM models, etc) by setting a style argument or passing in some standardized estimation function.

See also seaborn's tsplot function https://stanford.edu/~mwaskom/software/seaborn/generated/seaborn.tsplot.html and discussion on it's future: mwaskom/seaborn#896

define a class for plotting and fitting tuning functions: NeuroPop()

Parameters:

• self.X, n x 1, feature
• self.Y, n x m, population activity
• self.theta_, m x 1, preferred stimulus
• self.gain_, m x 1
• 'self.baseline_', m x 1

Methods:

• tunefit(self)
• plot(self)
• encode(self, X)
• decode(self,Y)
• poisslike(self,Y)