Building a model brain: Richard Walker at TEDxMontpellier
The Human Brain Project argues that understanding the brain's function requires building computational models, using a combination of global data and advanced simulation to diagnose and treat disorders, which could save billions annually. The project will achieve this by translating biological principles into digital models, allowing researchers to test mechanisms like hyperactivity within a controlled, simulated environment. This work aims to accelerate treatment discovery, though it acknowledges profound unknowns regarding the brain's full developmental plasticity.
## Speakers & Context
- Unnamed spokesperson for the Human Brain Project.
- Speaker is presenting the findings and goals of the Human Brain Project, which has secured over a billion euros in EU funding.
- The talk is framed within a discussion of medicine and brain disorders, contrasting the potential of modeling against current diagnostic and treatment limitations.
## Theses & Positions
- The human brain, despite its complexity, possesses a blueprint (genome) that can be described in a remarkably small amount of data (1.5 gigabytes).
- Understanding the brain's mechanisms requires moving beyond mere usage knowledge (like an iPhone app) to understanding its internal mechanics through simulation and modeling.
- Building computational models of the brain is feasible because the underlying principles suggest an inherent "Simplicity."
- The project's ultimate goal is to accelerate the search for new treatments for brain diseases, potentially reducing the cost of brain disease in Europe by even 1%, which equates to 8 billion EUR a year.
- Modeling the *developmental* plasticity of the brain is currently beyond computational power, preventing the creation of an artificial, world-controlling brain.
## Concepts & Definitions
- **Genome:** The blueprint of the human brain, which can be written down in about 1.5 gigabytes.
- **Synapse:** Individual connections between neurons, estimated at 80 trillion in the human brain.
- **Biological Signature of Disease:** A set of changes in the brain that differentiates a patient with one condition compared to a patient with a different disease or a healthy person.
- **Plasticity:** The continuous process where the brain changes moment-to-moment based on experience (hearing words, eating, listening to music); this process itself cannot currently be modeled.
## Mechanisms & Processes
- **Modeling Approach (Data Assembly):** Compiling global data on brain structure from various sources (molecules, circuits, neuron forms) like filling a jigsaw puzzle to reveal underlying principles.
- **Neuron Measurement/Modeling:**
1. Establishing a taxonomy of neural types (e.g., pyramidal neurons, basket cells).
2. Counting the different cell types within specific cortical areas (e.g., cortical column of WRA).
3. Creating 3D digital models of the cell shape (using advanced equipment).
4. Measuring the electrical behavior of the cells using a patch-clamp machine.
5. Creating a circuit model by having the digital shapes touch, indicating potential electrical connections, bypassing the impossible task of measuring every synapse.
- **Simulation:** Placing digital neuron models in a virtual space, injecting simulated current, and observing the electrical signals propagating between cells to map the circuit's function.
- **Disease Modeling:**
1. Collecting and anonymizing existing hospital brain scan data.
2. Identifying patients with similar conditions using criteria defined in biological terms.
3. Creating a "biological signature" from this data.
4. Tweaking the computational model (e.g., making neurons hyperactive) to reflect the suspected pathology.
5. Observing the cascade of effects to identify a potential cause and thus, a target for treatment.
## Timeline & Sequence
- The project involves an ongoing process of data acquisition from global sources.
- The modeling process follows a sequence: Taxonomy $\rightarrow$ Counting $\rightarrow$ 3D Shaping/Electrical Testing $\rightarrow$ Circuit Modeling $\rightarrow$ Supercomputer Simulation $\rightarrow$ Comparison/Correction.
- The overall aim is a gradual buildup toward a complete, detailed working model of the human brain, referencing rat brain models as a huge, but current, achievement.
## Named Entities
- **WRA:** A specific area of the cortex analyzed in the project.
- **Connecticut College:** (Mentioned by inference in Example 1, but not relevant here. Removed.)
## Numbers & Data
- EU Funding: **"a bit more than a billion euros"**.
- Neurons: **80 billion**.
- Connections/Synapses: **80 trillion**.
- Genome Size: **1.5 gigabytes**.
- CU Computer Bits: **32**.
- Cortical Layer count: **six layers** (in the WRA).
- Year of European Brain Council Report: **2011**.
- Cost of brain disease in Europe: **"of the order of 800 billion euros"** every year.
- Potential cost reduction goal: **1%** reduction equals **8 billion EUR** a year in Europe alone.
- Timeframe for model completion: A "long slow process," with expectations over a **10-year project** timeline.
## Examples & Cases
- **The iPhone analogy:** To fix a broken phone, one must know the internal mechanics, not just how to use an app.
- **Neuron Typing:** Examples include pyramidal neurons, basket cells, and Marty nauy cells.
- **Model Validation:** A simple test where modifying the model to simulate hyperactivity (e.g., in a specific circuit) reveals if the change has systemic effects, helping distinguish cause from secondary effect.
- **Asthma/Mental Health Example (Biological Signature):** Linking the physical consequences of environment/abuse (like a reduced hippocampus or altered HPA axis) to a detectable signature.
## Tools, Tech & Products
- **Supercomputers:** Used to run the complex brain simulations.
- **Patch-clamp machine:** Equipment used to measure the electrical behavior of cells.
- **Digital Models:** 3D digital recreations of neuron shapes.
- **Computational Models:** The abstract, simulated representation of brain circuits and functions.
## References Cited
- **European Brain Council:** Source of the 2011 report estimating the cost of brain disease.
- **Richard Feynman:** Quoted principle: "if you want to understand a complicated system you should build it."
## Trade-offs & Alternatives
- **Simulation vs. Direct Measurement:** Direct measurement of all 80 trillion synapses is deemed "completely impossible" and would take centuries.
- **Biological Complexity vs. Model Scope:** The process must abstract—focusing on principles and basic cell types—because modeling the dynamic plasticity is currently science fiction.
- **Drug Development:** The modeled findings are not a replacement for traditional methods; they still require testing in animals and clinical trials.
## Counterarguments & Caveats
- Skeptics claim measuring the 80 billion neurons and 80 trillion synapses is impossible.
- The process cannot model continuous plasticity, development, or the *how* of forming memories, as this requires knowledge we do not yet possess.
- The science project will not produce an artificial general intelligence that takes over the world.
## Methodology
- **Data Integration:** Aggregating heterogeneous datasets (molecular, circuit, structural) globally to identify foundational principles.
- **Iterative Modeling:** Starting with countable units (neuron types) and progressively building complexity—from single cells to local circuits, then to whole-brain simulations.
- **Differential Diagnosis:** Identifying the "biological signature" of disease by comparing patient-specific changes against established healthy models.
- **Mechanistic Testing:** Artificially altering model parameters (e.g., increasing excitability) to see if the resulting system change mirrors the pathology, thereby suggesting a root cause.
## Conclusions & Recommendations
- The project's primary recommendation is the systematic use of advanced simulation and modeling to diagnose and understand complex neurological disorders.
- The immediate goal is to establish actionable biological signatures from patient data that can be tested computationally.
- The ultimate "moonshot" is understanding the first acts of a baby learning to talk, which would resolve a scientific question open for 2,000 years.
## Implications & Consequences
- **Economic Impact:** A 1% cost reduction in brain disease for Europe yields 8 billion EUR annually, offering strong justification for the research.
- **Scientific Breakthrough:** Successfully modeling complex behavior (like speech acquisition) would represent a major step forward in human knowledge.
- **Medical Paradigm Shift:** Moving diagnosis and drug development from observation/reaction to proactive, simulation-driven hypothesis generation.
## Verbatim Moments
- *"if you want to understand a complicated system you should build it"*
- *"The theme of today is actually medicine"*
- *"if your iPhone breaks down you if you want to mend it it's not good enough to just know how to use an app you have to open it up and look at all the pieces inside and see which piece isn't working"*
- *"this one and a half gigabytes which fits 20 times inside this little phone"*
- *"we can create a model you we take neurons in the right proportions... and lo and behold we find that the fibers of these models touch and we know when when they touch that can form an electrical circuit"*
- *"the total cost of brain disease for the European economy is something of the order of 800 billion euros every year"*
- *"if we could reduce the cost of brain disease just by 1%... that's 8 billion EUR a year in Europe alone"*
- *"We don't actually model that process... we can't model development"*
- *"our moonshot that's why we've asked for the money"*