Classification of cortical areas using gene expression profiles
-
1
Boston University, Bioinformatics, United States
-
2
Boston University, Health Sciences, United States
-
3
Boston University, Speech, Language, and Hearing Sciences, United States
Brain atlases depict the parcellation of neural tissue into a set of spatially contiguous regions with distinct attributes. These “maps” traditionally reflect macroscopic anatomical or cytoarchitectural features and are only distantly connected to the proteins and molecules expressed in those areas. The Allen Brain Atlas (ABA) provides a collection of neuroanatomically-linked transcriptomic data collected with high spatial resolution at genome-scale. Determining the extent to which gene expression profiles differentiate the brain areas depicted in classical brain atlases begins to form a bridge between molecular anatomy and anatomical and functional brain organization. Several previous studies using the mouse ABA have demonstrated clustering of gene expression profiles that largely respect anatomical region boundaries (e.g., Lein et al., 2007; Bohland et al., 2010; Ko et al., 2013). Within the cortex, such clusters tend to follow laminar boundaries (e.g., Belgard et al., 2011), but some evidence supports more limited differential expression across cortical areas (e.g., Ng et al., 2009; Bohland et al., 2010).
Here we studied cortical expression profiles from the ABA using grid-based expression data (200 micron) registered to a 3D template. We used feature selection methods to choose most informative genes, and coupled these methods with support vector machines to learn relationships between normalized gene expression profiles and cortical region labels. We demonstrate results for classifiers trained to discriminate pairs of areas, which can achieve near 100% accuracy, and multi-class models that must classify any sample into one of 18 cortical regions. Using surprisingly few genes, a sample can be classified with > 70% accuracy. Feature selection methods achieve small but consistent improvements relative to random gene selection. To test whether classification accuracy is due to spatial autocorrelation independent of region boundaries, performance of classifiers trained on the reference atlas was compared to classifiers trained on random spatial parcellations of the cortex, constrained to match the distribution of region sizes in the reference atlas. While the latter achieved surprising accuracy, performance on the reference atlas was consistently better. Our results show that, while gene expression is relatively homogenous across the cortex, there are consistent transcriptomic differences that may underlie specialization of these regions.
Acknowledgements
This work was funded by a grant from the Dudley Allen Sargent Research Fund (J.W. Bohland, PI).
References
Belgard, T. G., et al. (2011). A transcriptomic atlas of mouse neocortical layers. Neuron, 71(4), 605-616.
Bohland, J. W., et al. (2010). Clustering of spatial gene expression patterns in the mouse brain and comparison with classical neuroanatomy. Methods, 50(2), 105-112.
Ko, Y., et al. (2013). Cell type-specific genes show striking and distinct patterns of spatial expression in the mouse brain. Proceedings of the National Academy of Sciences, 110(8), 3095-3100.
Lein, E. S., et al. (2007). Genome-wide atlas of gene expression in the adult mouse brain. Nature, 445(7124), 168-176.
Ng, L., et al. (2009). An anatomic gene expression atlas of the adult mouse brain. Nature Neuroscience, 12(3), 356-362.
Keywords:
Cerebral Cortex,
Gene Expression,
machine learning,
brain atlases,
bioinformatics
Conference:
Neuroinformatics 2014, Leiden, Netherlands, 25 Aug - 27 Aug, 2014.
Presentation Type:
Poster, to be considered for oral presentation
Topic:
Genomics and genetics
Citation:
Yan
R and
Bohland
JW
(2014). Classification of cortical areas using gene expression profiles.
Front. Neuroinform.
Conference Abstract:
Neuroinformatics 2014.
doi: 10.3389/conf.fninf.2014.18.00014
Copyright:
The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers.
They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.
The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.
Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.
For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.
Received:
04 Apr 2014;
Published Online:
04 Jun 2014.
*
Correspondence:
Prof. Jason W Bohland, Boston University, Health Sciences, Boston, MA, 02215, United States, j.bohland@pitt.edu