Biology subjects

Marin-Llobet, A.

Publications and source records attributed to Marin-Llobet, A..

6 recordsLinked to original sources

bioRxiv2025

DeviceAgent: An autonomous multimodal AI agent for flexible bioelectronics

The development of flexible bioelectronics remains a complex, multidisciplinary process that demands specialized expertise and labor-intensive efforts, limiting scalability, adaptability and accessibility. Here, we introduce DeviceAgent, an autonomous multimodal AI agent that integrates large language models (LLMs), vision-language models (VLMs), and domain-specific computational tools into a unified framework for bioelectronics research. Leveraging the emergent reasoning abilities of LLMs and VLMs, DeviceAgent enables zero and few-shot generalization, contextual learning, and flexible task execution across modalities. A multimodal context memory system orchestrates these capabilities, providing end-to-end support across the experimental pipeline-from high-level design objectives to fabrication protocol generation, visual defect inspection, and electrophysiological signal analysis, while maintaining human oversight at critical decision points. We demonstrate its capabilities through the development of stretchable mesh electronics for interfacing with human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs), a representative application involving complex device architectures, heterogeneous material nanofabrication, and electrophysiology analysis. DeviceAgent autonomously (1) generates customized bioelectronic layouts; (2) creates comprehensive fabrication protocols tailored to specific materials and processes; (3) identifies microscopic defects using visual reasoning; and (4) analyzes cardiac electrophysiological recordings in an interpretable manner. By embedding LLMs and VLMs within a structured, tool-augmented architecture, DeviceAgent establishes a scalable and accessible paradigm for AI-scientist collaboration in nanofabrication and bioelectronics research.

bioengineering↗

bioRxiv2025

An AI Agent for cell-type specific brain computer interfaces

Decoding how specific neuronal subtypes contribute to brain function requires linking extracellular electrophysiological features to underlying molecular identities, yet reliable in vivo electrophysiological signal classification remains a major challenge for neuroscience and clinical brain-computer interfaces (BCI). Here, we show that pretrained, general-purpose vision-language models (VLMs) can be repurposed as few-shot learners to classify neuronal cell types directly from electrophysiological features, without task-specific fine-tuning. Validated against optogenetically tagged datasets, this approach enables robust and generalizable subtype inference with minimal supervision. Building on this capability, we developed the BCI AI Agent (BCI-Agent), an autonomous AI framework that integrates vision-based cell-type inference, stable neuron tracking, and automated molecular atlas validation with real-time literature synthesis. BCI-Agent addresses three critical challenges for in vivo electrophysiology: (1) accurate, training-free cell-type classification; (2) automated cross-validation of predictions using molecular atlas references and peer-reviewed literature; and (3) embedding molecular identities within stable, low-dimensional neural manifolds for dynamic decoding. In rodent motor-learning tasks, BCI-Agent revealed stable, cell-type-specific neural trajectories across time that uncover previously inaccessible dimensions of neural computation. Additionally, when applied to human Neuropixels recordings-where direct ground-truth labeling is inherently unavailable-BCI-Agent inferred neuronal subtypes and validated them through integration with human single-cell atlases and literature. By enabling scalable, cell-type-specific inference of in vivo electrophysiology, BCI-Agent provides a new approach for dissecting the contributions of distinct neuronal populations to brain function and dysfunction.

neuroscience↗

bioRxiv2025

An autonomous AI agent for universal behavior analysis

Behavior analysis across species represents a fundamental challenge in neuroscience, psychology, and ethology, typically requiring extensive expert knowledge and labor-intensive processes that limit research scalability and accessibility. We introduce BehaveAgent, an autonomous multimodal AI agent designed to automate behavior analysis from video input without retraining or manual intervention. Unlike conventional methods that require manual behavior annotation, video segmentation, task-specific model training, BehaveAgent leverages the reasoning capabilities of multimodal large language models (LLM) to generalize across novel behavioral domains without need for additional training. It integrates LLMs, vision-language models (VLMs), and large-scale visual grounding modules, orchestrated through a multimodal context memory and goal-directed attention mechanism, to enable robust zero-shot visual reasoning across species and experimental paradigms, including plants, insects, rodents, primates, and humans. Upon receiving a video input, BehaveAgent autonomously identifies the correct analysis strategy and performs end-to-end behavior analysis and interpretation without human supervision. Leveraging vision-language representations, it performs general-purpose tracking, pose estimation and segmentation. We demonstrate BehaveAgents universal applicability to autonomously (1) identify the behavioral paradigm and develop an action plan specialized for the identified paradigm, (2) identify relevant subjects and objects, (3) track those features, (4) identify behavioral sequences with explicit reasoning, (5) generate and execute code for targeted analysis and (6) generate comprehensive research reports that integrate behavioral findings with relevant scientific literature. Through interpretable agentic reasoning, BehaveAgent makes its internal decision-making process transparent, clarifying why particular features are tracked or behaviors inferred. By reducing the time and expertise required for behavior analysis, BehaveAgent introduces a scalable, generalizable, and explainable paradigm for advancing biological and behavioral research.

neuroscience↗

bioRxiv2025

Spatial transcriptomics AI agent charts hPSC-pancreas maturation in vivo

Spatial transcriptomics has revolutionized our understanding of tissue organization by simultaneously capturing gene expression and spatial localization within intact tissues. However, analyzing these increasingly complex datasets requires specialized expertise across computational biology, statistics, and biological context. To address this challenge, we introduce the Spatial Transcriptomics AI Agent (STAgent), an autonomous multimodal agentic AI that integrates multimodal large language models (LLMs) with specialized computational tools to transform weeks-long analysis tasks into minutes of automated processing. Unlike conventional machine learning approaches that are limited to narrow, predefined tasks, STAgent leverages the emergent capabilities of multimodal LLMs - such as flexible reasoning, contextual understanding, and cross-modal integration - which allow it to adapt to novel data, execute multi-step analyses, and generate biologically meaningful insights with minimal human input. STAgent enables autonomous deep research through integrated capabilities, including dynamic code generation for complex analytical workflows, visual reasoning for interpreting spatial patterns, real-time retrieval of relevant peer-reviewd scientific literature, and synthesis of comprehensive, actionable reports. We applied STAgent to investigate the in vivo maturation of human stem cell-derived pancreatic cells (SC-pancreas) transplanted into immunodeficient mice. We generated single-cell spatial transcriptomics data spanning multiple developmental timepoints. STAgent autonomously (1) identified the maturation of initially scattered endocrine cells into well-defined islet-like structures, with predominantly peripheral -cells surrounding {beta}-cell cores supported by an expanding mesenchymal network; (2) revealed strengthening endocrine-endocrine cell interactions over time and, through context-aware gene set analysis, uncovered spatially resolved biological processes driving maturation; (3) unlike traditional analytical approaches, STAgent offers mechanistic explanations of spatial patterns, contextualizing findings with relevant literatures and developing cohesive insights into human pancreatic development. This agentic approach establishes a new paradigm in spatial transcriptomics analysis by substantially lowering the expertise barrier and reducing analysis time, accelerating biological and biomedical discovery.

genomics↗

bioRxiv2025

Spike sorting AI agent

Spike sorting is a fundamental process for decoding neural activity, involving preprocessing, spike detection, feature extraction, clustering, and validation. However, conventional spike sorting methods are highly fragmented, labor-intensive, and heavily reliant on expert manual curation, limiting their scalability and reproducibility. This challenge has become more pressing with advances in neural recording technology, such as high-density Neuropixels for large-scale neural recording or flexible electrodes for long-term stable recording over months to years. The volume and complexity of these datasets make manual curation infeasible, requiring an automated and scalable solution. Here, we introduce SpikeAgent, a multimodal large language model (LLM)-based AI agent that automates and standardizes the entire spike sorting pipeline. Unlike traditional approaches, SpikeAgent integrates multiple LLM backends, coding functions, and established algorithms, autonomously performing spike sorting with reasoning-based decision-making and real-time interaction with intermediate results. It generates interpretable reports, providing transparent justifications for each sorting decision, enhancing transparency and reliability. We benchmarked SpikeAgent against human experts across various neural recording technology, demonstrating its versatility and ability to achieve curation consistency that are equal to, or even higher than human experts. It also drastically reduces the expertise barrier and accelerates the curation and validation time by orders of magnitude. Moreover, it enables automated interpretability of the neural spiking data, which cannot be achieved by any conventional methods. SpikeAgent presents a paradigm shift in processing signals for neuroscience and brain-computer interfaces, while laying the ground for AI agent-augmented science across various domains.

neuroscience↗

bioRxiv2024

End-to-end multimodal deep learning for real-time decoding of months-long neural activity from the same cells

Long-term, stable, and real-time decoding of behavior-dependent neural dynamics from the same cells is critical for brain-computer interfaces (BCIs) and for understanding neural evolution during learning, and memory, and disease progression. Recent advances in flexible and high-density electrodes have enabled the stability required for long-term tracking but generate vast datasets that challenge existing analysis methods. Current spike sorting approaches rely heavily on manual curation and lack scalability for large-scale, real-time processing. Here, we introduce AutoSort, an end-to-end multimodal deep neural network-based method that enables real-time tracking and decoding of the same neurons over months. AutoSort uses a scalable strategy by learning deep representations from initial recordings and applying the trained model in real-time. It integrates multimodal features, including waveform features, distribution patterns, and inferred neuron spatial locations, to ensure robustness and accuracy. AutoSort outperforms existing methods in both simulated and long-term recordings, reducing computational demands by using only 10% of the time and 25% of the memory compared to conventional methods. By combining AutoSort with high-density flexible probes, we track neural dynamics in real-time during motor learning and skill acquisition over 2 months, capturing intrinsic neural manifold drift, stabilization, and post-learning representational drift. AutoSort offers a promising solution for studying long-term neural intrinsic dynamics and enabling real-time BCI decoding.

bioengineering↗