Comparative analysis of spike sorting using Uniform Manifold Approximation and Projection (UMAP) vs. feature-based methods in in vivo extracellular recordings.
<p>Recordings from Buzsáki’s lab [<a href="http://www.plosbiology.org/article/info:doi/10.1371/journal.pbio.3003527#pbio.3003527.ref026" target="_blank">26</a>] include an intracellular electrode serving as ground truth (GT, green) and four extracellular electro...
में बचाया:
| मुख्य लेखक: | |
|---|---|
| अन्य लेखक: | , , , , , , , , , , |
| प्रकाशित: |
2025
|
| विषय: | |
| टैग: |
टैग जोड़ें
कोई टैग नहीं, इस रिकॉर्ड को टैग करने वाले पहले व्यक्ति बनें!
|
| _version_ | 1849927641743753216 |
|---|---|
| author | Daniel Suárez-Barrera (22676654) |
| author2 | Lucas Bayones (22676657) Norberto Encinas-Rodríguez (22676660) Sergio Parra (22676663) Viktor Monroy (22676666) Sebastián Pujalte (22676669) Bernardo Andrade-Ortega (22676672) Héctor Díaz (22676675) Manuel Alvarez (3468647) Antonio Zainos (22676678) Alessio Franci (143351) Román Rossi-Pool (22676681) |
| author2_role | author author author author author author author author author author author |
| author_facet | Daniel Suárez-Barrera (22676654) Lucas Bayones (22676657) Norberto Encinas-Rodríguez (22676660) Sergio Parra (22676663) Viktor Monroy (22676666) Sebastián Pujalte (22676669) Bernardo Andrade-Ortega (22676672) Héctor Díaz (22676675) Manuel Alvarez (3468647) Antonio Zainos (22676678) Alessio Franci (143351) Román Rossi-Pool (22676681) |
| author_role | author |
| dc.creator.none.fl_str_mv | Daniel Suárez-Barrera (22676654) Lucas Bayones (22676657) Norberto Encinas-Rodríguez (22676660) Sergio Parra (22676663) Viktor Monroy (22676666) Sebastián Pujalte (22676669) Bernardo Andrade-Ortega (22676672) Héctor Díaz (22676675) Manuel Alvarez (3468647) Antonio Zainos (22676678) Alessio Franci (143351) Román Rossi-Pool (22676681) |
| dc.date.none.fl_str_mv | 2025-11-24T18:33:37Z |
| dc.identifier.none.fl_str_mv | 10.1371/journal.pbio.3003527.g006 |
| dc.relation.none.fl_str_mv | https://figshare.com/articles/figure/Comparative_analysis_of_spike_sorting_using_Uniform_Manifold_Approximation_and_Projection_UMAP_vs_feature-based_methods_in_in_vivo_extracellular_recordings_/30697592 |
| dc.rights.none.fl_str_mv | CC BY 4.0 info:eu-repo/semantics/openAccess |
| dc.subject.none.fl_str_mv | Cell Biology Molecular Biology Neuroscience Physiology Biotechnology Developmental Biology Science Policy Space Science Biological Sciences not elsewhere classified Mathematical Sciences not elsewhere classified Information Systems not elsewhere classified spike sorting pipelines spike sorting pipeline seldom spiking neurons recorded putative neurons correctly sorted neurons clustering algorithms responsible spike sorting makes reliable spike sorting umap drastically increases data analysis techniques drastically improve processed data experimental data universal practice precise explorations neural recordings neural code mathematically grounded fundamentally motivated extracellular electrophysiology enables deeper crucial step computational cost ad hoc |
| dc.title.none.fl_str_mv | Comparative analysis of spike sorting using Uniform Manifold Approximation and Projection (UMAP) vs. feature-based methods in in vivo extracellular recordings. |
| dc.type.none.fl_str_mv | Image Figure info:eu-repo/semantics/publishedVersion image |
| description | <p>Recordings from Buzsáki’s lab [<a href="http://www.plosbiology.org/article/info:doi/10.1371/journal.pbio.3003527#pbio.3003527.ref026" target="_blank">26</a>] include an intracellular electrode serving as ground truth (GT, green) and four extracellular electrodes (<i>E</i><sub><i>1</i></sub><i>–E</i><sub><i>4</i></sub>) placed at different distances from the GT neuron. <b>(A)</b> UMAP-based sorting on electrode <i>E</i><sub><i>1</i></sub> identifies four clusters (distinct shapes). One cluster (<i>S</i><sub><i>E1</i></sub>, circular markers) overlaps substantially with the GT neuron, while the other three represent separate putative neurons. <b>(B)</b> Raster plots for GT clusters found by UMAP on each extracellular electrode, alongside the GT neuron’s raster. Each raster line is horizontally split; the color saturation of the top half represents the precision (percentage of cluster spikes that come from the GT), and the color saturation of the bottom half indicates the recall (percentage of GT spikes captured by that cluster) (see color bar in G). Electrodes <i>E</i><sub><i>1</i></sub> and <i>E</i><sub><i>2</i></sub> align well with the GT neuron (uniform dark blue), whereas <i>E</i><sub><i>3</i></sub>’s cluster is contaminated (low precision), and <i>E</i><sub><i>4</i></sub>’s cluster loses many GT spikes (low recall). <b>(C)</b> Inclusion Index matrix <i>M</i><sub><i>i,j</i></sub> for UMAP, illustrating overlap of GT spikes across electrode clusters. Darker squares for <i>E</i><sub><i>1</i></sub> and <i>E</i><sub><i>2</i></sub> reflect accurate capture of GT spikes, while <i>E</i><sub><i>3</i></sub> and <i>E</i><sub><i>4</i></sub> show contamination and loss, respectively. <b>(D)</b> UMAP-sorted waveforms from <i>E</i><sub><i>1</i></sub>, color-coded to match the clusters in (A). <b>(E)</b> Experimental schematic from Buzsáki’s lab, showing the intracellular GT neuron and four extracellular electrodes. <b>(F)</b> UMAP sorting on electrode <i>E</i><sub><i>3</i></sub> highlights significant contamination, with many not-GT spikes mislabeled as GT (shown in a separate color). <b>(G)</b> Color bar for the inclusion matrices (C, J, M). <b>(H–J)</b> Wavelet-based sorting: (H) projection of spikes from <i>E</i><sub><i>1</i></sub> in Wavelet feature space, identifying multiple clusters. (I) Raster plots for the GT clusters on each electrode. The GT neuron’s pattern is not well isolated. (J) Inclusion matrix shows low, inconsistent overlap for GT spikes across electrodes. <b>(K–M)</b> PCA-based sorting: (K) PCA projection for <i>E</i><sub><i>1</i></sub>, resulting in broad, overlapping clusters. (L) Raster plots for these GT clusters showing substantial contamination. (M) Inclusion matrix reflects the poor specificity of PCA-based clustering, with larger fractions of non-GT in each putative cluster. The data containing simultaneous intracellular and extracellular recordings from the hippocampus can be found at [<a href="http://www.plosbiology.org/article/info:doi/10.1371/journal.pbio.3003527#pbio.3003527.ref040" target="_blank">40</a>], and the analysis code is available at [<a href="http://www.plosbiology.org/article/info:doi/10.1371/journal.pbio.3003527#pbio.3003527.ref049" target="_blank">49</a>].</p> |
| eu_rights_str_mv | openAccess |
| id | Manara_064cb59fb01c30a244e03fdf1914750a |
| identifier_str_mv | 10.1371/journal.pbio.3003527.g006 |
| network_acronym_str | Manara |
| network_name_str | ManaraRepo |
| oai_identifier_str | oai:figshare.com:article/30697592 |
| publishDate | 2025 |
| repository.mail.fl_str_mv | |
| repository.name.fl_str_mv | |
| repository_id_str | |
| rights_invalid_str_mv | CC BY 4.0 |
| spelling | Comparative analysis of spike sorting using Uniform Manifold Approximation and Projection (UMAP) vs. feature-based methods in in vivo extracellular recordings.Daniel Suárez-Barrera (22676654)Lucas Bayones (22676657)Norberto Encinas-Rodríguez (22676660)Sergio Parra (22676663)Viktor Monroy (22676666)Sebastián Pujalte (22676669)Bernardo Andrade-Ortega (22676672)Héctor Díaz (22676675)Manuel Alvarez (3468647)Antonio Zainos (22676678)Alessio Franci (143351)Román Rossi-Pool (22676681)Cell BiologyMolecular BiologyNeurosciencePhysiologyBiotechnologyDevelopmental BiologyScience PolicySpace ScienceBiological Sciences not elsewhere classifiedMathematical Sciences not elsewhere classifiedInformation Systems not elsewhere classifiedspike sorting pipelinesspike sorting pipelineseldom spiking neuronsrecorded putative neuronscorrectly sorted neuronsclustering algorithms responsiblespike sorting makesreliable spike sortingumap drastically increasesdata analysis techniquesdrastically improveprocessed dataexperimental datauniversal practiceprecise explorationsneural recordingsneural codemathematically groundedfundamentally motivatedextracellular electrophysiologyenables deepercrucial stepcomputational costad hoc<p>Recordings from Buzsáki’s lab [<a href="http://www.plosbiology.org/article/info:doi/10.1371/journal.pbio.3003527#pbio.3003527.ref026" target="_blank">26</a>] include an intracellular electrode serving as ground truth (GT, green) and four extracellular electrodes (<i>E</i><sub><i>1</i></sub><i>–E</i><sub><i>4</i></sub>) placed at different distances from the GT neuron. <b>(A)</b> UMAP-based sorting on electrode <i>E</i><sub><i>1</i></sub> identifies four clusters (distinct shapes). One cluster (<i>S</i><sub><i>E1</i></sub>, circular markers) overlaps substantially with the GT neuron, while the other three represent separate putative neurons. <b>(B)</b> Raster plots for GT clusters found by UMAP on each extracellular electrode, alongside the GT neuron’s raster. Each raster line is horizontally split; the color saturation of the top half represents the precision (percentage of cluster spikes that come from the GT), and the color saturation of the bottom half indicates the recall (percentage of GT spikes captured by that cluster) (see color bar in G). Electrodes <i>E</i><sub><i>1</i></sub> and <i>E</i><sub><i>2</i></sub> align well with the GT neuron (uniform dark blue), whereas <i>E</i><sub><i>3</i></sub>’s cluster is contaminated (low precision), and <i>E</i><sub><i>4</i></sub>’s cluster loses many GT spikes (low recall). <b>(C)</b> Inclusion Index matrix <i>M</i><sub><i>i,j</i></sub> for UMAP, illustrating overlap of GT spikes across electrode clusters. Darker squares for <i>E</i><sub><i>1</i></sub> and <i>E</i><sub><i>2</i></sub> reflect accurate capture of GT spikes, while <i>E</i><sub><i>3</i></sub> and <i>E</i><sub><i>4</i></sub> show contamination and loss, respectively. <b>(D)</b> UMAP-sorted waveforms from <i>E</i><sub><i>1</i></sub>, color-coded to match the clusters in (A). <b>(E)</b> Experimental schematic from Buzsáki’s lab, showing the intracellular GT neuron and four extracellular electrodes. <b>(F)</b> UMAP sorting on electrode <i>E</i><sub><i>3</i></sub> highlights significant contamination, with many not-GT spikes mislabeled as GT (shown in a separate color). <b>(G)</b> Color bar for the inclusion matrices (C, J, M). <b>(H–J)</b> Wavelet-based sorting: (H) projection of spikes from <i>E</i><sub><i>1</i></sub> in Wavelet feature space, identifying multiple clusters. (I) Raster plots for the GT clusters on each electrode. The GT neuron’s pattern is not well isolated. (J) Inclusion matrix shows low, inconsistent overlap for GT spikes across electrodes. <b>(K–M)</b> PCA-based sorting: (K) PCA projection for <i>E</i><sub><i>1</i></sub>, resulting in broad, overlapping clusters. (L) Raster plots for these GT clusters showing substantial contamination. (M) Inclusion matrix reflects the poor specificity of PCA-based clustering, with larger fractions of non-GT in each putative cluster. The data containing simultaneous intracellular and extracellular recordings from the hippocampus can be found at [<a href="http://www.plosbiology.org/article/info:doi/10.1371/journal.pbio.3003527#pbio.3003527.ref040" target="_blank">40</a>], and the analysis code is available at [<a href="http://www.plosbiology.org/article/info:doi/10.1371/journal.pbio.3003527#pbio.3003527.ref049" target="_blank">49</a>].</p>2025-11-24T18:33:37ZImageFigureinfo:eu-repo/semantics/publishedVersionimage10.1371/journal.pbio.3003527.g006https://figshare.com/articles/figure/Comparative_analysis_of_spike_sorting_using_Uniform_Manifold_Approximation_and_Projection_UMAP_vs_feature-based_methods_in_in_vivo_extracellular_recordings_/30697592CC BY 4.0info:eu-repo/semantics/openAccessoai:figshare.com:article/306975922025-11-24T18:33:37Z |
| spellingShingle | Comparative analysis of spike sorting using Uniform Manifold Approximation and Projection (UMAP) vs. feature-based methods in in vivo extracellular recordings. Daniel Suárez-Barrera (22676654) Cell Biology Molecular Biology Neuroscience Physiology Biotechnology Developmental Biology Science Policy Space Science Biological Sciences not elsewhere classified Mathematical Sciences not elsewhere classified Information Systems not elsewhere classified spike sorting pipelines spike sorting pipeline seldom spiking neurons recorded putative neurons correctly sorted neurons clustering algorithms responsible spike sorting makes reliable spike sorting umap drastically increases data analysis techniques drastically improve processed data experimental data universal practice precise explorations neural recordings neural code mathematically grounded fundamentally motivated extracellular electrophysiology enables deeper crucial step computational cost ad hoc |
| status_str | publishedVersion |
| title | Comparative analysis of spike sorting using Uniform Manifold Approximation and Projection (UMAP) vs. feature-based methods in in vivo extracellular recordings. |
| title_full | Comparative analysis of spike sorting using Uniform Manifold Approximation and Projection (UMAP) vs. feature-based methods in in vivo extracellular recordings. |
| title_fullStr | Comparative analysis of spike sorting using Uniform Manifold Approximation and Projection (UMAP) vs. feature-based methods in in vivo extracellular recordings. |
| title_full_unstemmed | Comparative analysis of spike sorting using Uniform Manifold Approximation and Projection (UMAP) vs. feature-based methods in in vivo extracellular recordings. |
| title_short | Comparative analysis of spike sorting using Uniform Manifold Approximation and Projection (UMAP) vs. feature-based methods in in vivo extracellular recordings. |
| title_sort | Comparative analysis of spike sorting using Uniform Manifold Approximation and Projection (UMAP) vs. feature-based methods in in vivo extracellular recordings. |
| topic | Cell Biology Molecular Biology Neuroscience Physiology Biotechnology Developmental Biology Science Policy Space Science Biological Sciences not elsewhere classified Mathematical Sciences not elsewhere classified Information Systems not elsewhere classified spike sorting pipelines spike sorting pipeline seldom spiking neurons recorded putative neurons correctly sorted neurons clustering algorithms responsible spike sorting makes reliable spike sorting umap drastically increases data analysis techniques drastically improve processed data experimental data universal practice precise explorations neural recordings neural code mathematically grounded fundamentally motivated extracellular electrophysiology enables deeper crucial step computational cost ad hoc |