iSVM: an incremental machine learning approach for recreating ion channel profiles in neocortical neurons
-
1
Ecole Polytechnique Fédérale de Lausanne, Brain Mind Institute, Switzerland
Purpose: Voltage-gated ion channels play an important role in determining the intrinsic firing properties of neurons by regulating the flow of ions and controlling the voltage gradient across the neuronal membrane. Genetic studies have identified nearly 200 different monomeric and heteromeric ion channels expressed throughout the brain. Experimental observations suggest that different sets of ion channels could underlie the same morpho-electrical subtypes. Understanding this molecular diversity is a fundamental goal in neuroscience. Here, we present an incremental Support Vector Machine (iSVM) based model that estimates the expression of ion channel genes in different morpho-electric neurons with a high level of accuracy.
Method: Gene expressions of 135 neurons were profiled for 26 voltage-gated ion channels using single cell RT-PCR measurements and were categorized into three feature classes according to neuron properties: layer, morphology class and electrical firing type. We first identified which channels have a significant change in expression between the classes and then used iSVM to build a predictive model for those channels. The iSVM model is initially trained and tuned using the three feature classes. We then increment the number of features by combining the expression of every gene to the three feature classes and recompute the prediction accuracy of the remaining genes. If the accuracy is improved, we retain the combined gene, otherwise, we reject it. We iteratively combine more genes until the prediction accuracy can no longer be improved.
Results: The results show that most of the channels have a significant change in expression between classes indicating that the layer, morphology, and electrical type of the neuron have an important relationship to the expression of ion channel genes. Although, the correlation coefficients between channels are less than 0.48, the iSVM model can significantly improve the prediction accuracy of some channels by more than 10% when taking into account the expression of other channels. However, the prediction of channels for which the expression frequency was less than 10% could not be improved. Using a 10 fold cross-validation test, the iSVM model obtains an overall average accuracy greater than 83% as opposed to 67% obtained when using logistic regression. Additionally, iSVM was able to recreate the ion channel expression of a test dataset that was not used while building the model with 76% average accuracy.
Conclusion: These results show that it is possible to predict the ion channel diversity for a range of neuronal subtypes in the rat neocortex for which gene expression data has not yet been gathered.
Conference:
Neuroinformatics 2010 , Kobe, Japan, 30 Aug - 1 Sep, 2010.
Presentation Type:
Poster Presentation
Topic:
General neuroinformatics
Citation:
Khazen
G,
Hill
S,
Schürmann
F and
Markram
H
(2010). iSVM: an incremental machine learning approach for recreating ion channel profiles in neocortical neurons.
Front. Neurosci.
Conference Abstract:
Neuroinformatics 2010 .
doi: 10.3389/conf.fnins.2010.13.00051
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:
11 Jun 2010;
Published Online:
11 Jun 2010.
*
Correspondence:
Georges Khazen, Ecole Polytechnique Fédérale de Lausanne, Brain Mind Institute, Lausanne, Switzerland, gkhazen@lau.edu.lb