Research Using IBM Cloud and AI Sheds Light on Huntington’s Disease (HD)
New research combining the power of brain modeling, IBM’s AI, and the IBM Cloud sheds light on how brain cells interact in Huntington’s disease (HD) and could lead to a better understanding of how these interactions impact other brain diseases as well. IBM’s Healthcare and Life Sciences team collaborated with the CHDI Foundation and researchers at the University of California, Los Angeles (UCLA) on the study, whose results are published today in Cell Reports.
We launched our collaboration with the CHDI Foundation, a nonprofit biomedical research organization devoted to understanding HD, in April 2015. Our ongoing work together seeks to decipher the causes of HD progression, ranging from how networks of cells within the brain function to how a single gene could cause the brain to degenerate so tragically in HD.
Much of our research focuses on the striatum, a structure deep within the brain and known for motor coordination. The striatum is the first element of the brain to succumb to this disease. Years before the onset of disease symptoms, cells there begin to die1. Shedding light on key dysfunctions within this structure is important for a number of reasons, from uncovering clues about the root causes of HD to defining new therapies to prevent this devastating disease from taking hold.
New breakthroughs relevant to both our collaboration with CHDI and our understanding of the chain of events that lead to HD have today been published in Cell Reports. Our models of a confounding optogenetic experiment might shed light on how brain cells interact pathologically in HD. We worked closely with CHDI and researchers at UCLA, using IBM Cloud-based modeling to demonstrate—for the first time—that certain light-sensitive proteins introduced into cells within the striatum can trigger the same reactions that are implicated in brain dysfunction in an HD model animal.
This new discovery not only provides subtle clues into how our brains function, but also signals loudly to the scientific community that current research methods—widely accepted as the gold standard in the field—could be unexpectedly altering these reactions and causing a misinterpretation of research results.
We started by building models of the principal cells in the striatum, known as medium spiny neurons (MSNs). Our models first captured how these neurons could have an excessive reaction to stimuli, or become “hyperexcitable”, and vulnerable to degeneration when the concentration of potassium throughout the medium in which they live is elevated, as it is in HD.
During our work, a study2 in Nature Neuroscience caught our eye, from the lab of Baljith (Bal) Khakh, Professor of Physiology and Neurobiology at the Geffen School of Medicine at UCLA. Bal is a leading researcher of astrocytes (a major type of “non-neuronal” cells in the brain) and this study demonstrated how a dysfunctional channel in astrocytes can make them unable to clear extracellular potassium, leading to neuronal hyperexcitability in transgenic mice that model HD.
Our team began collaborating with Bal and his team at UCLA about this connection, exploring whether changes in extracellular potassium could play a role in certain of their experiments. The team was using optogenetic tools, a common and standard experimental technique in neuroscience.
Optogenetics, which has been highlighted as a breakthrough and gold standard of brain measurement by leading scientific journals3, uses light to precisely control neurons. It is often used to probe and explore brain activity and study the behavior of neurons under excitation, largely because many believe it to have a laser-like precision.
The UCLA team shared the data from 14 experiments conducted with optogenetics, in which neurons were stimulated by blue flashes of a LED light and then found to have elevated levels of extracellular potassium. In turn, this led to consequential changes in their excitability, similar to reactions observed in HD. When the light was turned off, the potassium concentration fell back to baseline and the cells became less excitable. It was as if these experiments could cause an HD-related cellular response and then reverse it in just about 10 minutes.
These findings lay the foundation for understanding how light stimulations can alter the responses of neurons and other cells within the brain, create crosstalk, and potentially change the accuracy of experiments.
This new knowledge could not only guide neurophysiologists to better interpret experiments, but also help them to draw the connection to how miscommunication and interactions between these cells might be setting up the brain for the devastating effects of HD.
We were able to run these calculations and models thanks to the power of the IBM Cloud, which allowed us to compute thousands of brain cell interaction models, along with the empirical data obtained from past experiments by UCLA. We found 110,000 unique sets of neuron parameters that were able to reproduce the conditions of light-on and light-off experiments and then replicate the effects of the confounding potassium rise. We then used further optimizations to narrow these down to ~1,100 models, each of which closely matched the empirical data.
IBM’s brain modeling platform provides a quantitative framework to evaluate biological mechanisms at different spatial and temporal scales using mechanistic simulations, evolutionary algorithms and AI. A snapshot of the different neuron types and networks that undergo degeneration in Huntington’s disease that were studied using the IBM Neural Tissue Simulator are highlighted.
The impact of our models does not stop at HD. The novelty and uniqueness of the tools our team developed4, 5, 6 are allowing us to reproduce these cell populations under varying conditions, which could open the door to comprehending how these interactions could impact other brain diseases.
We’re beginning to explore the variation of brain cells across healthy, diseased, and drug-treated conditions, as well as the inner workings of dopamine neurons, which are central to addiction, Parkinson’s disease, and striatal neurons’ responses to HD. With the combined power of brain modeling, IBM’s AI, and the IBM Cloud, we hope this work will lead to even more significant discoveries about the inner workings of neurodegenerative disease and ultimately give clues into how to best prevent and treat them.
References
-
Ross et al., Nature Reviews Neurology volume 10, pages 204–216 (2014), “Huntington disease: natural history, biomarkers and prospects for therapeutics”. ↩
-
https://www.frontiersin.org/articles/10.3389/fninf.2011.00015/full ↩
-
http://www.abstractsonline.com/pp8/index.html#!/4376/presentation/8383 ↩