Sorted Spikes Decoding¶

The mechanics of decoding with sorted spikes are largely similar to those of decoding with unsorted spikes. You should familiarize yourself with the clusterless decoding tutorial before proceeding with this one.

The elements we will need to decode with sorted spikes are:

PositionGroup
SortedSpikesGroup
DecodingParameters
encoding_interval
decoding_interval

This time, instead of extracting waveform features, we can proceed directly from the SpikeSortingOutput table to specify which units we want to decode. The rest of the decoding process is the same as before.

In [6]:

Copied!





from pathlib import Path
import datajoint as dj

dj.config.load(
    Path("../dj_local_conf.json").absolute()
)  # load config for database connection info
from pathlib import Path
import datajoint as dj

dj.config.load(
    Path("../dj_local_conf.json").absolute()
)  # load config for database connection info

SortedSpikesGroup¶

SortedSpikesGroup is a child table of SpikeSortingOutput in the spikesorting pipeline. It allows us to group the spikesorting results from multiple sources (e.g. multiple tetrode groups or intervals) into a single entry. Here we will group together the spiking of multiple tetrode groups to use for decoding.

This table allows us filter units by their annotation labels from curation (e.g only include units labeled "good", exclude units labeled "noise") by defining parameters from UnitSelectionParams. When accessing data through SortedSpikesGroup the table will include only units with at least one label in include_labels and no labels in exclude_labels. We can look at those here:

In [7]:

Copied!





from spyglass.spikesorting.analysis.v1.group import UnitSelectionParams

UnitSelectionParams().insert_default()

# look at the filter set we'll use here
unit_filter_params_name = "default_exclusion"
print(
    (
        UnitSelectionParams()
        & {"unit_filter_params_name": unit_filter_params_name}
    ).fetch1()
)
# look at full table
UnitSelectionParams()
from spyglass.spikesorting.analysis.v1.group import UnitSelectionParams

UnitSelectionParams().insert_default()

# look at the filter set we'll use here
unit_filter_params_name = "default_exclusion"
print(
    (
        UnitSelectionParams()
        & {"unit_filter_params_name": unit_filter_params_name}
    ).fetch1()
)
# look at full table
UnitSelectionParams()

{'unit_filter_params_name': 'default_exclusion', 'include_labels': [], 'exclude_labels': ['noise', 'mua']}

Out[7]:

unit_filter_params_name	include_labels	exclude_labels
all_units	=BLOB=	=BLOB=
default_exclusion	=BLOB=	=BLOB=
exclude_noise	=BLOB=	=BLOB=
MS2220180629	=BLOB=	=BLOB=

Total: 4

Now we can make our sorted spikes group with this unit selection parameter

In [8]:

Copied!





from spyglass.spikesorting.spikesorting_merge import SpikeSortingOutput
import spyglass.spikesorting.v1 as sgs

nwb_copy_file_name = "mediumnwb20230802_.nwb"

sorter_keys = {
    "nwb_file_name": nwb_copy_file_name,
    "sorter": "mountainsort4",
    "curation_id": 1,
}
# check the set of sorting we'll use
(
    sgs.SpikeSortingSelection & sorter_keys
) * SpikeSortingOutput.CurationV1 & sorter_keys
from spyglass.spikesorting.spikesorting_merge import SpikeSortingOutput
import spyglass.spikesorting.v1 as sgs

nwb_copy_file_name = "mediumnwb20230802_.nwb"

sorter_keys = {
    "nwb_file_name": nwb_copy_file_name,
    "sorter": "mountainsort4",
    "curation_id": 1,
}
# check the set of sorting we'll use
(
    sgs.SpikeSortingSelection & sorter_keys
) * SpikeSortingOutput.CurationV1 & sorter_keys

Out[8]:

sorting_id	merge_id	recording_id	sorter	sorter_param_name	nwb_file_name name of the NWB file	interval_list_name descriptive name of this interval list	curation_id
642242ff-5f0e-45a2-bcc1-ca681f37b4a3	75286bf3-f876-4550-f235-321f2a7badef	01c5b8e9-933d-4f1e-9a5d-c494276edb3a	mountainsort4	franklab_tetrode_hippocampus_30KHz	mediumnwb20230802_.nwb	0a6611b3-c593-4900-a715-66bb1396940e	1
a4b5a94d-ba41-4634-92d0-1d31c9daa913	143dff79-3779-c0d2-46fe-7c5040404219	a8a1d29d-ffdf-4370-8b3d-909fef57f9d4	mountainsort4	franklab_tetrode_hippocampus_30KHz	mediumnwb20230802_.nwb	3d782852-a56b-4a9d-89ca-be9e1a15c957	1
874775be-df0f-4850-8f88-59ba1bbead89	a900c1c8-909d-e583-c377-e98c4f0deebf	747f4eea-6df3-422b-941e-b5aaad7ec607	mountainsort4	franklab_tetrode_hippocampus_30KHz	mediumnwb20230802_.nwb	9cf9e3cd-7115-4b59-a718-3633725d4738	1

Total: 3

Finding the merge id's corresponding to an interpretable restriction such as merge_id or interval_list can require several join steps with upstream tables. To simplify this process we can use the included helper function SpikeSortingOutput().get_restricted_merge_ids() to perform the necessary joins and return the matching merge id's

In [11]:

Copied!





# get the merge_ids for the selected sorting
spikesorting_merge_ids = SpikeSortingOutput().get_restricted_merge_ids(
    sorter_keys, restrict_by_artifact=False
)

# create a new sorted spikes group
unit_filter_params_name = "default_exclusion"
SortedSpikesGroup().create_group(
    group_name="test_group",
    nwb_file_name=nwb_copy_file_name,
    keys=[
        {"spikesorting_merge_id": merge_id}
        for merge_id in spikesorting_merge_ids
    ],
    unit_filter_params_name=unit_filter_params_name,
)
# check the new group
SortedSpikesGroup & {
    "nwb_file_name": nwb_copy_file_name,
    "sorted_spikes_group_name": "test_group",
}
# get the merge_ids for the selected sorting
spikesorting_merge_ids = SpikeSortingOutput().get_restricted_merge_ids(
    sorter_keys, restrict_by_artifact=False
)

# create a new sorted spikes group
unit_filter_params_name = "default_exclusion"
SortedSpikesGroup().create_group(
    group_name="test_group",
    nwb_file_name=nwb_copy_file_name,
    keys=[
        {"spikesorting_merge_id": merge_id}
        for merge_id in spikesorting_merge_ids
    ],
    unit_filter_params_name=unit_filter_params_name,
)
# check the new group
SortedSpikesGroup & {
    "nwb_file_name": nwb_copy_file_name,
    "sorted_spikes_group_name": "test_group",
}

Out[11]:

nwb_file_name name of the NWB file	unit_filter_params_name	sorted_spikes_group_name
mediumnwb20230802_.nwb	default_exclusion	test_group

Total: 1

In [8]:

Copied!





# look at the sorting within the group we just made
SortedSpikesGroup.Units & {
    "nwb_file_name": nwb_copy_file_name,
    "sorted_spikes_group_name": "test_group",
    "unit_filter_params_name": unit_filter_params_name,
}
# look at the sorting within the group we just made
SortedSpikesGroup.Units & {
    "nwb_file_name": nwb_copy_file_name,
    "sorted_spikes_group_name": "test_group",
    "unit_filter_params_name": unit_filter_params_name,
}

Out[8]:

nwb_file_name name of the NWB file	unit_filter_params_name	sorted_spikes_group_name	spikesorting_merge_id
mediumnwb20230802_.nwb	default_exclusion	test_group	143dff79-3779-c0d2-46fe-7c5040404219
mediumnwb20230802_.nwb	default_exclusion	test_group	75286bf3-f876-4550-f235-321f2a7badef
mediumnwb20230802_.nwb	default_exclusion	test_group	a900c1c8-909d-e583-c377-e98c4f0deebf

Total: 3

Model parameters¶

As before we can specify the model parameters. The only difference is that we will use the ContFragSortedSpikesClassifier instead of the ContFragClusterlessClassifier.

In [9]:

Copied!





from spyglass.decoding.v1.core import DecodingParameters
from non_local_detector.models import ContFragSortedSpikesClassifier


DecodingParameters.insert1(
    {
        "decoding_param_name": "contfrag_sorted",
        "decoding_params": ContFragSortedSpikesClassifier(),
        "decoding_kwargs": dict(),
    },
    skip_duplicates=True,
)

DecodingParameters()
from spyglass.decoding.v1.core import DecodingParameters
from non_local_detector.models import ContFragSortedSpikesClassifier


DecodingParameters.insert1(
    {
        "decoding_param_name": "contfrag_sorted",
        "decoding_params": ContFragSortedSpikesClassifier(),
        "decoding_kwargs": dict(),
    },
    skip_duplicates=True,
)

DecodingParameters()

Out[9]:

decoding_param_name a name for this set of parameters	decoding_params initialization parameters for model	decoding_kwargs additional keyword arguments
contfrag_clusterless	=BLOB=	=BLOB=
contfrag_clusterless_0.5.13	=BLOB=	=BLOB=
contfrag_clusterless_6track	=BLOB=	=BLOB=
contfrag_sorted	=BLOB=	=BLOB=
contfrag_sorted_0.5.13	=BLOB=	=BLOB=
j1620210710_contfrag_clusterless_1D	=BLOB=	=BLOB=
j1620210710_test_contfrag_clusterless	=BLOB=	=BLOB=
MS2220180629_contfrag_sorted	=BLOB=	=BLOB=
ms_lineartrack_2023_contfrag_sorted	=BLOB=	=BLOB=
ms_lineartrack_contfrag_clusterless	=BLOB=	=BLOB=
ms_lineartrack_contfrag_sorted	=BLOB=	=BLOB=
ms_wtrack_2023_contfrag_sorted	=BLOB=	=BLOB=

...

Total: 15

1D Decoding¶

As in the clusterless notebook, we can decode 1D position if we specify the track_graph, edge_order, and edge_spacing parameters in the Environment class constructor. See the clusterless decoding tutorial for more details.

Decoding¶

Now we can decode the position using the sorted spikes using the SortedSpikesDecodingSelection table. Here we assume that PositionGroup has been specified as in the clusterless decoding tutorial.

In [2]:

Copied!





selection_key = {
    "sorted_spikes_group_name": "test_group",
    "unit_filter_params_name": "default_exclusion",
    "position_group_name": "test_group",
    "decoding_param_name": "contfrag_sorted",
    "nwb_file_name": "mediumnwb20230802_.nwb",
    "encoding_interval": "pos 0 valid times",
    "decoding_interval": "test decoding interval",
    "estimate_decoding_params": False,
}

from spyglass.decoding import SortedSpikesDecodingSelection

SortedSpikesDecodingSelection.insert1(
    selection_key,
    skip_duplicates=True,
)
selection_key = {
    "sorted_spikes_group_name": "test_group",
    "unit_filter_params_name": "default_exclusion",
    "position_group_name": "test_group",
    "decoding_param_name": "contfrag_sorted",
    "nwb_file_name": "mediumnwb20230802_.nwb",
    "encoding_interval": "pos 0 valid times",
    "decoding_interval": "test decoding interval",
    "estimate_decoding_params": False,
}

from spyglass.decoding import SortedSpikesDecodingSelection

SortedSpikesDecodingSelection.insert1(
    selection_key,
    skip_duplicates=True,
)

In [14]:

Copied!

from spyglass.decoding.v1.sorted_spikes import SortedSpikesDecodingV1

SortedSpikesDecodingV1.populate(selection_key)
from spyglass.decoding.v1.sorted_spikes import SortedSpikesDecodingV1

SortedSpikesDecodingV1.populate(selection_key)

[12:19:30][WARNING] Spyglass: Upsampled position data, frame indices are invalid. Setting add_frame_ind=False
No GPU/TPU found, falling back to CPU. (Set TF_CPP_MIN_LOG_LEVEL=0 and rerun for more info.)

Encoding models:   0%|          | 0/54 [00:00<?, ?cell/s]

Non-Local Likelihood:   0%|          | 0/54 [00:00<?, ?cell/s]

We verify that the results have been inserted into the DecodingOutput merge table.

In [3]:

Copied!

from spyglass.decoding.decoding_merge import DecodingOutput

DecodingOutput.SortedSpikesDecodingV1 & selection_key
from spyglass.decoding.decoding_merge import DecodingOutput

DecodingOutput.SortedSpikesDecodingV1 & selection_key

Out[3]:

merge_id	nwb_file_name name of the NWB file	unit_filter_params_name	sorted_spikes_group_name	position_group_name	decoding_param_name a name for this set of parameters	encoding_interval descriptive name of this interval list	decoding_interval descriptive name of this interval list	estimate_decoding_params whether to estimate the decoding parameters
42e9e7f9-a6f2-9242-63ce-94228bc72743	mediumnwb20230802_.nwb	default_exclusion	test_group	test_group	contfrag_sorted	pos 0 valid times	test decoding interval	0

Total: 1

We can load the results as before:

In [ ]: