Summary: Recording the brain’s electrical “noise” has been possible for decades, but pinpointing which specific cell types are producing those signals has remained a major challenge. Researchers at Boston University have developed PhysMAP, a machine learning framework that distinguishes neuronal cell types from their unique electrophysiological signatures recorded in vivo.
PhysMAP can identify cell types implicated in psychiatric conditions—such as those linked to schizophrenia and major depressive disorder—directly from live electrical recordings. This capability opens new avenues for studying how particular neural circuits fail in real time and for designing targeted, next-generation therapies for circuit-based psychiatric disorders.
Key Facts
- Targeting circuitopathies: PhysMAP is aimed at disorders that arise from dysfunctional interactions between specific cell types rather than global changes in brain activity. Examples include schizophrenia, major depressive disorder, and some forms of epilepsy.
- Separating individual “voices”: The algorithm combines multiple electrophysiological features—such as spike shape, timing, and frequency content—to identify neuron classes like parvalbumin-positive and somatostatin-positive cells within mixed, multiunit recordings.
- No genetic tagging required: Unlike approaches that rely on optotagging or genetic manipulation, PhysMAP can infer cell identity from electrical recordings alone, enabling study of cell types in living brains without additional molecular interventions.
- Built from open data: The tool was trained and validated on seven public datasets that included both electrical recordings and verified cell-type labels, demonstrating how shared scientific data can be repurposed to develop new diagnostic and research tools.
Source: Boston University
When silicon probes are inserted into the brain for research or clinical monitoring, they capture the electrical activity of many neurons at once. Those signals reveal how neural circuits compute and, in some cases, where they break down. However, without knowing which cell types contribute to recorded patterns, it is difficult to link circuit dynamics to the underlying cellular mechanisms that drive healthy function or disease.
Brains are composed of many neuron classes that play distinct roles in computation and are differentially affected by psychiatric disorders and pharmacological agents. To design treatments that target malfunctioning circuits, researchers need tools that can resolve cell-type–specific activity within bulk electrophysiological recordings.
A multidisciplinary team at Boston University’s Chobanian & Avedisian School of Medicine, College of Arts & Sciences, College of Engineering, and Faculty of Computing & Data Sciences developed PhysMAP to address this need. The framework integrates multiple complementary electrophysiological features to create multimodal representations that reveal cell-type identity within complex recordings.
PhysMAP adapts methods from multiomics analysis to weight diverse electrophysiological modalities simultaneously, producing interpretable representations aligned with transcriptomic cell types. This multimodal approach performs better than any single feature alone and enables researchers to identify putative cell types even when ground-truth labels are not available.
The team trained and tested PhysMAP using seven publicly available datasets that paired single-neuron electrical recordings with verified cell-type identities obtained via optotagging—an approach that uses light-based molecular tools to confirm cell identity. Once PhysMAP learned the mapping between electrical signatures and cell types, the mapping could be transferred to new, untagged datasets, allowing simultaneous study of multiple cell types without further molecular tagging.
This work also highlights the power of open data sharing: publicly released datasets allowed the development and validation of a new tool without the need for additional experiments. The researchers report that PhysMAP can identify cell types implicated in psychiatric disorders, including parvalbumin-positive interneurons (relevant to schizophrenia and Dravet syndrome) and somatostatin-positive interneurons (implicated in major depressive disorder).
“A growing number of psychiatric disorders appear to arise from disrupted interactions between specific cell types—so-called circuitopathies—rather than simply global increases or decreases in activity,” says Chandramouli (Chand) Chandrasekaran, PhD, corresponding author and assistant professor of anatomy & neurobiology and psychological and brain sciences at Boston University. “PhysMAP makes it possible to study these interacting cell types in vivo, which could inform development of future therapeutic strategies.”
A prior iteration of this approach, WaveMAP, was used to identify cell types in the first human Neuropixels recordings. PhysMAP extends that capability and offers greater power to distinguish multiple clinically relevant cell types from high-density electrode data.
The study and its results are published in Nature Communications.
Funding: CC was supported by NIH NINDS R00NS092972, R01NS121409, R21NS135361, and R01NS122969; the Moorman-Simon Interdisciplinary Career Development Professorship at Boston University; the Whitehall Foundation (2019-12-77); and the Brain and Behavior Research Foundation Young Investigator Award (27923). Additional dataset- and investigator-specific support included NIH grants listed in the original report.
Key Questions Answered:
A: Overall activity is like the roar of a stadium—you hear the volume but not who is leading the chant. Many psychiatric and neurological disorders arise when specific neuron classes fail to coordinate properly. Identifying the contributing cell types reveals the circuit-level sources of dysfunction and points to more precise interventions.
A: PhysMAP was trained on optotagged datasets where light-based methods confirmed cell identity. The algorithm learned characteristic patterns—timing, waveform shape, firing dynamics—associated with each cell class and can then recognize those signatures in untagged recordings.
A: That is the long-term aim. Because PhysMAP does not require genetic manipulation, it has the potential to work with clinical electrode recordings (including high-density probes such as Neuropixels) to help clinicians infer the cellular origins of symptoms and guide more targeted treatments.
Editorial Notes:
- This article was edited by a Neuroscience News editor.
- The journal paper was reviewed in full for this summary.
- Additional context was added by editorial staff.
About this AI mental health research news
Author: Gina DiGravio
Source: Boston University
Contact: Gina DiGravio – Boston University
Image: Image credit: Neuroscience News
Original Research: Open access. “A multimodal approach for visualizing and identifying electrophysiological cell types in vivo” by Eric Kenji Lee, Asım E. Gül, Greggory Heller, Anna Lakunina, Han Yu, Andrew Shelton, Shawn Olsen, Nicholas A. Steinmetz, Cole Hurwitz, Santiago Jaramillo, Pawel F. Przytycki & Chandramouli Chandrasekaran. DOI: 10.1038/s41467-026-71331-0
Abstract
A multimodal approach for visualizing and identifying electrophysiological cell types in vivo
Different neuron types perform distinct computations and coordinate activity during sensation, perception, and action. Electrophysiological recordings can capture many neurons simultaneously, but identifying cell types within these recordings has been challenging.
PhysMAP is a framework that integrates multiple electrophysiological modalities to produce interpretable, multimodal representations. Applied to seven datasets, PhysMAP’s multimodal representations align more closely with transcriptomically defined cell types than any single modality alone. This alignment enables better identification of putative cell types when ground-truth labels are absent and allows annotated datasets to transfer labels to unannotated recordings. PhysMAP also helps detect batch effects that can confound classification. Collectively, these results establish PhysMAP as a tool for studying multiple cell types simultaneously and for gaining insight into neural circuit dynamics.