Some forecasting on the next decade of Neuroscience
Based on The BRAIN Initiative 2025 Report and the BRAIN Initiative August 2025 Meeting
The NIH BRAIN Initiative marked its 10th anniversary with a report on its impacts in systems and computational neuroscience and team-scale research, as well as its August 2025 meeting.
The Initiative was originally framed by identifying 7 priority areas: (1) Discovering Diversity, (2) Maps at Multiple Scales, (3) The Brain in Action, (4) Demonstrating Causality, (5) Identifying Fundamental Principles, (6) Advancing Human Neuroscience, and (7) From the BRAIN Initiative to the Brain.
Over the past decade, the Initiative has focused on a) quantifying complex behavior and simultaneous brain recordings, b) supporting the creation of tools and databases for tractable large-scale data, and c) contributing to changes in research culture by normalizing team science, cross-disciplinary training and neuroethics.
At its meeting, the Initiative set out a roadmap for the next ten years organized around four pillars: BRAIN Knowledgebase, Precision molecular circuit therapies, Accelerating human neuroscience, and NeuroAI.
The Knowledgebase aims to overcome data fragmentation by enabling large-scale analysis and reuse. Precision molecular circuit therapies focus on modulating dysfunctional brain circuits safely and effectively. Accelerating human neuroscience seeks to de-risk innovation and enable targeted clinical applications. NeuroAI will build brain-inspired, ethically grounded predictive models that learn and adapt.
The BRAIN Initiative predicts the following accomplishments in the next 10 years (quote):
Plug-and-play, user-friendly, and secure AI-powered BRAIN knowledge ecosystem
Therapeutic platforms that safely modulate neural circuits with molecular precision, enabling targeted treatment of a range of chronic brain conditions
Clear translational pathways from foundational knowledge and de-risked neurotechnologies to therapeutic interventions in real-world settings
Fundamental principles of brain function that reciprocally inform natural intelligence and artificial intelligence responsibly and without bias
Brain-inspired technological applications for adaptable, secure, resilient, and energ-yefficient health monitoring and interventions
Hence, four axes for the next decade:
1. New Techniques and Technologies. Tools such as digital twins to model brain systems (for clinical care, drug testing, and decision-making), next-generation neural recording and modulation at cellular and millisecond scales, and synchronized behavioral quantification platforms that align rich human and animal data.
2. Focus Areas. Priorities are linking discoveries to mental health and brain disorders, advancing precision circuit therapies, expanding human neuroscience datasets, pushing further integration of computational modeling, and embedding safeguards for neural data.
3. Methods and Analytical approaches. Deeper integration of machine learning, both for analyzing multimodal datasets and for developing biologically inspired AI architectures. Coupling theories and generative models with experimental validation from the get-go, to construct healthy databases. Cross-modal data integration through interoperable databases, along with standardized data practices, to ensure reproducibility.
4. Scales of investigation. Research will continue across molecular → cellular → circuit → systems → behavioral levels, with attention to temporal scales (from ms to longitudinal) and population scales. Cross-species comparisons for identifying conserved computations. Human neuroscience benefits from anticipated shared clinical and experimental datasets of higher precision.
Some remarks :
The BRAIN Initiative’s philosophy is to understand the brain as a complex system that enables adaptive interaction with the environment. Because all neurological, mental, and behavioral disorders arise from system-level dysfunctions, understanding these dynamics is crucial for prevention and treatment.
First, the NeuroAI direction is obviously salient. Current AI does not work the way the brain does; advancing the field requires new architectures and models informed by biological systems. In turn, insights from neuroscience inform the design of AI, while AI tools will support analysis, prediction, and generative modeling in brain science.
Second, though NeuroAI grows directly out of computational neuroscience, computational neuroscience has always circulated in both directions: from the brain to AI (using biological circuit principles to inspire artificial systems) and from AI to brain (using machine learning and generative models to analyze and interpret neural data). In the same sense, NeuroAI focused on mental health disorders closely intersect with computational psychiatry. 2025-2035 renews the direction of the computational cognitive sciences program.
Third, several challenges accompany the roadmap.
Scientifically, priorities include accounting for multi-level phenotyping, aka modelling individual variability, bridging molecular and circuit levels, and validating models against experimental data. This requires strong dimentionality reduction.
Research will likely rely increasingly on existing large-scale datasets (such as the ABCD study) rather than constant new data collection, yet adoption remains limited. We don’t know yet why.
Differences between sectors complicate alignment across sectors and obscure funding opportunities, such as the industry focusing on neuromodulation (especially underwhelming headsets & portable EEG) versus neural circuitry in research.
The programs to come will require building new tools, technologies and infrastructures. Bridging public and private interests, clarifying access to clinical data, and diversifying funding models (venture capital, non-profits, clinical partnerships) will require coordination, and especially strong guardrails.
Thanks to new technologies, we are growing the scale of human neural recordings in both quantity and precision. That demands infrastructure and standards.
Best practices will involve coupling model building with model testing, and leveraging cross-species datasets to identify conserved principles of computation across the lifespan. Best practices already tend towards computational reproducibility; digital twins will likely become part of them as well.