Researchers from the Computational Intelligence Group based at the Universidad Politécnica de Madrid's Facultad de Informática have used machine learning and data mining techniques to compare gene expresssion levels in Alzheimer's disease (AD) patients in two key regions of the hippocampus: the dentate gyrus, where the disease appears to have little or no effect, and the entorhinal cortex, where Alzheimer's disease produces major neuronal damage. The results, published in Computer Methods and Programs in Biomedicine, corroborate previous findings by other studies and set forth new working hypotheses for AD research.
Dentate gyrus and entorhinal cortex
The hippocampal formation is a complex structure situated in the medial temporal lobe of the brain. It plays a key role in memory, attention and spatial navigation. It is composed of six regions: presubiculum, subiculum, parasubiculum, dentate gyrus, hippocampus and entorhinal cortex. A noteworthy part of this formation is the dentate gyrus, one of the few regions of the brain where new neurons continue to be born (neurogenesis) throughout adulthood. It is known that neurogenesis plays a key role in the generation of new memories. Postmortem histopathological studies have provided evidence that the dentate gyrus is the hippocampal region least affected by AD.
The entorhinal cortex is the primary interface between the hippocampus and the neocortex. This region plays a key role in memory formation and consolidation, as well as the retrieval of autobiographical, declarative and spatial memories. Today, we know that the entorhinal cortex is not only one of the first areas to be affected by AD but also where its progression produces most lesions.
Data analysis using an ensemble of Bayesian classifiers
The data mining technique used is part of the computational intelligence discipline. Thanks to the computational power available today, vast amounts of complex data can be analysed holistically to identify new findings or set forth new hypotheses. This study has used ensemble statistical techniques applied to mathematical models to search for relevant genes and gene dependency networks. In both cases, the mathematical paradigm used is called Bayesian classifier.
Results