Neurogenetic profiles of human memory components

On August 10, 2015, Proceedings of the National Academy of Sciences published a paper titled “Computational dissection of human episodic memory reveals mental process-specific genetic profiles”  We have asked one of the authors of this research, Dr Gediminas Luksys from the University of Basel, to comment on this work.

The Study

Human memory is one of the most complex and essential cognitive traits. Although its intricacies have been studied well by psychologists and neuroscientists, its genetic underpinnings are less well known. Twin studies have shown its substantial heritability and hypothesis-based studies as well as genome-wide scans discovered a number of genetic markers associated with it. However, most of these studies considered memory as a whole, employing general phenotypes such as numbers of items people recall, whereas it has been known that memory performance depends on distinct mental processes that often correspond to distinct neural mechanisms.

In our study we applied computational modeling to investigate processes of human episodic memory, such as learning and memory maintenance, as well as factors modulating it, such as positive and negative emotional arousal. We also took into account that decision-making processes, like willingness to guess under uncertainty, influence memory performance. As different model parameters signified different memory-related processes, we estimated them from episodic memory performance of 1765 healthy young adults to construct each individual’s memory profile. Then we looked for sets of biologically related genes that were enriched in associations with these model parameters. Our results linked genetic variations in amine transporters, neural cell adhesion molecules, and collagens to memory encoding, maintenance, and its modulation by negative emotional arousal, respectively.

We also investigated neural mechanisms underlying these differences. Using functional neuroimaging we found that large clusters of brain activity differences in frontal cortical areas were mediating the association between neural cell adhesion molecules and memory maintenance. Overall, our study showed, for the first time, that computational models could be helpful in uncovering the genetic basis of specific memory processes that are closer to neurobiology and hence more likely to be addressable by therapeutic interventions than generic memory scores.

Background

The Brain Simulates Paths to a Brighter Future Hippocampal place cells, which create an internal map during novel exploration, are getting ‘replayed’ during sleep

Computational models are developed to improve our understanding of how our brain works at different levels: neuronal, systems, cognitive. Whereas neuronal models have been closely linked to biological data for many years, for cognitive models their relation to biology has been more vague. However, a number of key studies in the previous decade showed that particularly for complex cognitive processes such as reinforcement learning, memory and decision-making, model variables and parameters can be closer related to neurobiology than externally observable behaviors. Some studies also showed that taking into account individual variability one could understand the effects of different environmental factors (e.g. stress) on cognition and behavior much better.

As the field of human psychiatric genetics has been rapidly expanding during the recent years, we figured that to ensure its success it was critical to concentrate not just on ever expanding sample sizes and complex genetic approaches but also on the other end of the equation that remained remarkably, often overly simple – behavioral phenotypes. Computational model-based analysis is an important development in this direction. It allows for a more specific characterization of neurocognitive processes of interest than generic phenotypes do. Although models can be customized for different tasks, because of their fundamental nature certain model parameters and variables can generalize between tasks much better than usual behavioral performance measures do.

Future Direction

The findings of the present study contribute to a better understanding of the complex processes of human memory and we hope that they will eventually lead to the development of new and more targeted treatment options for various memory disorders in the future. In addition, studies of human cognition and behavior slowly move out of the laboratory domain and with the help of Internet and crowdsourcing more realistic and powerful alternatives come into play, where computational model-based analysis may prove to be an efficient tool. Finally, the model-based approach allows testing various therapeutic manipulations in silico before they are tested experimentally, which may prove to be much more cost effective.