Advanced Computational and Imaging Tools Yield Insights Into Learning and Decision-Making

Research aims to extend observations of reversal learning in mice to human neurological disorders


Cutting-edge optical imaging techniques, advanced computational methods and complex analysis of animal behavior are being employed to better understand — and even manipulate — neurological mechanisms underlying learning and decision-making. Using preclinical mouse models of Alzheimer’s disease and autism, Murat Yildirim, PhD, assistant staff in Cleveland Clinic’s Department of Neurosciences, is conducting innovative investigations designed to help develop new diagnostic and therapeutic strategies for a variety of neurological disorders.


Cleveland Clinic is a non-profit academic medical center. Advertising on our site helps support our mission. We do not endorse non-Cleveland Clinic products or services. Policy

“We are creating and employing tools applying engineering, optics, photonics and behavioral and computational neuroscience to precisely correlate behavioral deficits with abnormalities in brain activity,” says Dr. Yildirim. “We are further experimenting with optogenetic techniques on a cellular level to correct behavioral deficits.”

Reversal learning investigations: a jumping-off point

Dr. Yildirim’s work in this realm involves reversal learning experimentation, an important investigative method in neurobiology in animal models as well as humans. In animal models, this involves rewarding a behavior such as spinning a wheel in a certain direction and then changing the rules to give rewards only for the wheel going in the opposite direction.

In general, responses to reversal learning can be characterized as follows:

  1. Naïve subjects have an initial random trial-and-error period during which they make many errors as they figure out the system. As they gain expertise, errors are typically substantially reduced.
  2. When the direction for rewards is abruptly changed, some subjects can quickly learn the new system and keep errors to a minimum. This “inference-based learning” indicates understanding of the game.
  3. Some subjects persist in a suboptimal strategy referred to as Q learning, model-free learning or reinforcement learning. Subjects may get better at the task but continue to make multiple errors.

While neurotypical people tend to quickly learn the most efficient way to maximize rewards in reversal learning situations, many with neurological disorders — including Alzheimer’s disease, Parkinson’s disease, schizophrenia and autism — tend to be less adept.

Before joining Cleveland Clinic in 2022, Dr. Yildirim was part of an investigative team at Massachusetts Institute of Technology that used complex statistics to characterize in detail how mice behave under reversal learning conditions. The research was recently described in PLOS Computational Biology (2023 Sep;19[9]:e1011430).


The team subjected 21 healthy mice to a reversal learning task involving wheel spinning, as described above. The researchers quantified and analyzed mice behavior over multiple trials using the block Hidden Markov Model, a computational method that accounts for dynamic factors inherent in task learning. They found that the vast majority of mice, although they apparently learned the game and improved their rewards beyond chance, continued to employ a mixture of strategies rather than a single optimal one.

The investigators speculated that while it is possible that mice never fully develop inference-based learning because they are less capable of learning effectively or are forgetful, they may have adaptive reasons for persisting in what — in this experimental setting — is a suboptimal strategy.

“In a natural environment, with changing food sources and the ever-present threat of predators, mice are likely better off if they continue to explore, even after they have found what appears to be a secure food supply,” says Dr. Yildirim, who was a study co-author. “Our analytic tools allowed us to observe that mouse behavior is not as uniform as many results published in the literature lead us to believe.”

Deeper explorations with preclinical models of neurological disease

Dr. Yildirim is currently applying analytic techniques used in the earlier investigations as well as innovative optogenetic tools involving fluorescent and label-free imaging to study mouse preclinical models of Alzheimer’s disease and autism. Like humans with these disorders, the mice are less likely than controls to transition from Q learning to inference-based strategies.

These methods allow investigators to not only record but create perturbations of neural circuits using light therapy, which results in complex behavior in vitro (using cerebral organoids, i.e., miniature brain-like organs created artificially from stem cells) as well as in vivo.


To further these investigations, Dr. Yildirim’s lab is developing next-generation multiphoton systems, enabling superior imaging of living biological tissues. This imaging, which can focus down to the cellular level, is minimally invasive and can be conducted deep in tissues over long time periods, enabling characterization of dynamic processes without damage.

“We are currently examining brain mechanisms underlying differences in mice behavioral strategies in reversal learning situations and assessing whether we can manipulate these strategies,” Dr. Yildirim explains.

Translational research will be next

Dr. Yildirim soon hopes to collaborate with clinical researchers to pursue investigations in humans.

“With rapid advances in functional MRI, functional near-infrared spectroscopy and transcranial magnetic stimulation, the ability to precisely treat behavioral disorders without medications may become reality,” he predicts. “The basic research we are doing in developing and applying advanced technologies to mouse models can provide an important basis for moving this field forward.”

Related Articles

gut microbes in intestine
Cleveland Clinic, Tufts University Research Ties Gut Microbial TMAO Pathway to Chronic Kidney Disease

Large-scale joint study links elevated TMAO blood levels and chronic kidney disease risk over time

patient in ICU
Cleveland Clinic and Purdue Seek to Revolutionize Intensive Care Through AI

Investigators are developing a deep learning model to predict health outcomes in ICUs.

Multi-Ancestry Genetic Study of Parkinson’s Disease Identifies New Risk Genes in Pursuit of Novel Treatment Targets

International collaboration is most genetically diverse study of the disease to date

Noninvasive Technology Enhances Ability to Map Brain Activity to Track Behavior Change

Preclinical work promises large-scale data with minimal bias to inform development of clinical tests

Can Boosting Hydrogen Sulfide Bolster Standard-of-Care Glioblastoma Therapy to Extend Survival?

Cleveland Clinic researchers pursue answers on basic science and clinical fronts

Microglial Immunometabolism Endophenotypes Implicated in Sex Differences in Alzheimer’s Disease

Study suggests sex-specific pathways show potential for sex-specific therapeutic approaches

23-CCC-4375928 Quantum Innovation Catalyzer 650×450
A Unique Opportunity to Explore Quantum Computing’s Potential

Cleveland Clinic launches Quantum Innovation Catalyzer Program to help start-up companies access advanced research technology

Light trails coming from African American’s head
Blood-Based Biomarkers for Alzheimer’s Disease in Women (Podcast)

Research project aims to pinpoint biomarkers that could speed diagnosis